[PDF page 1] FSPS PUB 51 FEDERAL INFORMATION PROCESSING STANDARDS PUBLICATION 1974 JUNE U.S. DEPARTMENT OF COMMERCE / National Bureau of Standards Guidelines FOR AUTOMATIC DATA PROCESSING PHYSICAL SECURITY AND RISK MANAGEMENT CATEGORY: ADR OPERATIONS SUBCATEGORY: COMPUTER SECURITY [PDF page 2] Foreword The Federal Information Processing Standards Publication Series of the National Bureau of Standards is the official publication relating to standards adopted and promul¬ gated under the provisions of Public Law 89-306 (Brooks Bill) and under Part 6 of Title 15, Code of Federal Regulations. These legislative and executive mandates have given the Secretary of Commerce important responsibilities for improving the utilization and man¬ agement of computers and automatic data processing systems in the Federal Government. To carry out the Secretary’s responsibilities, the NBS, through its Institute for Computer Sciences and Technology, provides leadership, technical guidance, and coordination of government efforts in the development of guidelines and standards in these areas. The subject areas of personal privacy, data confidentiality and computer security are of the greatest national interest. The Secretary of Commerce has identified the efforts required to provide solutions to technical problems encountered in these areas as personal objectives in the Department’s overall program. Data confidentiality and computer security are dependent upon the application of a balanced set of managerial and technological safeguards. Within the context of a total security program, the NBS is pleased to make these Guidelines for ADP Physical Security and Risk Management available for use by Federal agencies. Ruth M. Davis, Director Institute for Computer Sciences and Technology) Abstract This publication provides guidelines to be used by Federal organizations in struc¬ turing physical security programs for their ADP facilities. It treats security analysis, natural disasters, supporting utilities, system reliability, procedural measures and controls, off-site facilities, contingency plans, security awareness and security audit. It contains statistics and information relevant to physical security of computer data and facilities and references many applicable publications for a more exhaustive treat¬ ment of specific subjects. tKeywords: ADP security; computer reliability; contingency plans; Federal Information Processing Standard; fire safety, natural disasters; physical security; risk analysis; security audit; security awareness; supporting utilities. Nat. Bur. Stand. (U.S.), Fed. Info. Process. Stand. Publ. (FIPS PUB) 31, 92 pages, (1974) CODEN: FIPPAT For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington D -0402. (Order by SL> Catalog C 13..>2:31) GPO, price $1.35 Subscription service also availal for all new FIPS publications and supplements for an indefinite period. C. ble [PDF page 3] FIPS PUB 31 Federal Information Processing Standards Publication 31 1974 June ANNOUNCING THE GUIDELINES FOR AUTOMATIC DATA PROCESSING PHYSICAL SECURITY AND RISK MANAGEMENT Federal Information Processing Standards Publications are issued by the National Bureau of Standards pursuant to the Federal Property and Administrative Services Act of 1949 as amended, Public Law 89-306 (79 Stat. 1127), and as implemented by Executive Order 11717 (38 FR 12315, dated May 11, 1973), and Part 6 of Title 15 CFR (Code of Federal Regulations). Name of Standard. Guidelines for Automatic Data Processing Physical Security and Risk Man¬ agement. Category of Standard. ADP Operations, Computer Security. Explanation: These guidelines provide a handbook for use by Federal organizations in structuring- physical security and risk management programs for their ADP facilities. This publication dis¬ cusses security analysis, natural disasters, supporting utilities, system reliability, procedural meas¬ ures and controls, off-site facilities, contingency plans, security awareness, and security audit. It contains statistics and information relevant to physical security of computer data and facilities and references many applicable publications for a more exhaustive treatment of specific subjects. Approving Authority. Department of Commerce, National Bureau of Standards (Institute for Computer Sciences and Technology). Maintenance Agency. Department of Commerce, National Bureau of Standards (Institute for Com¬ puter Sciences and Technology). Cross Index. None. Applicability. These Guidelines are intended as basic reference document and a checklist for gen¬ eral use throughout the Federal Government to evaluate computer security and plan physical se¬ curity programs in ADP systems. Implementation. As new ADP systems are developed and current systems improved, these Guide¬ lines should be utilized. Each organization should analyze its requirements for protection of data and processing facilities and implement the recommendations found in these Guidelines commen¬ surate to its calculated risk. Depending upon differing operational requirements, facilities will re¬ quire various levels of security protection. These Guidelines should assist the installation of man¬ agers in making, and justifying essential security decisions. Specifications. Federal Information Processing Standard 31 (FIPS 31), Guidelines for Automatic Data Processing Physical Security and Risk Management, (affixed). Qualifications. The statistics and recommendations provided in these Guidelines are based upon data and information supplied from many sources within the government and private sectors and reflect current practice and technologies. As new knowledge, techniques, and equipments become available in the future, these Guidelines will need to be modified accordingly. As experiences are gained through use and application of these Guidelines, a basis for security standards may be es¬ tablished. In this regard, comments and critiques concerning applications experience will be wel¬ comed. These should be addressed to the Associate Director for ADP Standards, Institute for Com¬ puter Sciences and Technology, National Bureau of Standards, Washington, D.C. 20234. 1 [PDF page 4] FIPS PUB 31 Where to Obtain Copies of the Standard. a. Copies of this publication are available from the Superintendent of Documents, U.S. Govern¬ ment Printing Office, Washington, D.C. 20402 (SD Catalog Number 03.52:31). There is a 25 per¬ cent discount on quantities of 100 or more. When ordering, specify document number, title, and SD Catalog Number. Payment may be made by check, money order, coupons, or deposit account. b. Microfiche of this publication is available from the National Technical Information Service, U.S. Department of Commerce, Springfield, Virginia 22151. When ordering refer to Report Num¬ ber NBS-FIPS-PUB-31 and title. Payment may be made by check, money order, coupons, or deposit account. 2 [PDF page 5] FIPS PUB 31 Federal Information Processing Standards Publication 31 1974 June Specifications for GUIDELINES FOR AUTOMATIC DATA PROCESSING PHYSICAL SECURITY AND RISK MANAGEMENT Contents Page Action Summary 5 1. ADP Security Analysis 8 1.0 Introduction 8 1.1 Scope 8 1.2 Threat to ADP Operations 9 1.3 Risk Analysis . 9 1.3.1 Loss Potential Estimate 9 1.3.2 Threat Analysis 11 1.3.3 Annual Loss Expectancy 11'' 1.3.4 Selecting- Remedial Measures 13 1.4 Implementing the Security Program 14 1.5 Supporting Documents 15 2. Anticipating Natural Disasters 15 2.0 Introduction 15 2.1 Fire Safety 15 2.1.1 ADP Facility Fire Exposure 16 2.1.2 Fire Detection 17 2.1.3 Fire Extinguishment 18 2.1.4 Fire Fighting 20 2.2 Flood 21 2.3 Earthquake 23 2.4 Windstorms 23 3. Supporting Utilities 27 3.0 Introduction 27 3.1 Electric Power 27 3.2 Air Conditioning 34 3.3 Communications Circuit 39 3.4 Other Supporting Utilities 42 4. Computer System Reliability 42 4.0 Introduction . 42 4.1 Computer System Reliability 42 4.2 Management of Hardware Maintenance 43 4.3 Reliability Considerations for New Systems 44 5. Physical Protection of ADP Facilities 45 5.0 Introduction 45 5.1 Determining Protection Requirements 45 Page 5.1.1 Instructions for the Facility Physical Security Survey 46 5.2 Boundary Protection 47 5.2.1 Emanations 48 5.3 Entrance Door Controls 48 5.4 Perimeter Intrusion Controls 49 5.5 Critical Area Controls 50 5.6 Guard Force Operations 53 5.7 Integrating Physical Security Measures 54 6. Internal Controls 54 6.0 Introduction 54 6.1 Personnel Controls 55 6.1.1 Personnel Selection 55 6.1.2 Training 55 6.1.3 Supervision 55 6.2 Organizing for Internal Control 56 6.3 Data Controls 59 6.4 Data Retention and Back-Up 59 6.4.1 Short Term Back-Up 59 6.4.2 Long Term Back-Up 59 6.5 Programming Controls 60 6.5.1 Program Design 60 6.5.2 Program Installation 60 6.5.3 Documentation of Controls 62 7. Security of Off-Site ADP Facilities 62 7.0 Introduction 62 7.1 Analysis of Security Requirements 63 7.2 On-Site Security 63 7.3 In-Transit Security 64 7.4 Off-Site Security 64 8. Contingency Planning 65 8.0 Introduction 65 8.1 Preparation of Contingency Plans 65 8.2 Emergency Response Planning 68 8.3 Back-Up Operations Planning 69 8.4 Recovery Planning 71 8.5 Testing Contingency Plans 73 9. Security Awareness and Communications 73 9.0 Introduction 73 3 [PDF page 6] FIPS PUB 31 Page Page 9.1 Senior Management 74 10.3 Conducting the Audit 81 9.2 Communicating the Security Program 74 10.4 Follow-up 82 9.2.1 Target Audience for the ADP Security Plan 74 Appendices 9.2.2 Content of Communication Plan 74 9.2.3 Method of Communication 74 A. Glossary 83 9.3 Summary 75 B. Bibliography 84 10. Internal Audit of Physical Security 75 C. Sample Table of Contents of Programming 10.0 Introduction 75 Procedures 86 10.1 Audit Preparation 76 10.2 The Audit Plan 76 Index 91 4 [PDF page 7] FIPS PUB 31 Acknowledgments The Institute for Computer Sciences and Technology acknowledges Robert V. Jacob¬ son, Vice President of SENTOR Security Group, Inc., as the principal author of this publication, and Dr. William F. Brown of Ball State University and Peter S. Browne of General Electric Information Systems as contributing authors. The work was done under contract to the Systems and Software Division of the Institute. The Institute wishes to thank the Office of Federal Protective Services Management, General Services Administration and the Federal Fire Council, as well as the Safety and Fire Protection Section of NBS and the NBS Fire Protection Service for their scrutiny of pertinent chapters and suggestions for modification. The Institute is grateful to Alfred M. Pfaff, Research Associate sponsored by the IBM Corporation, for reviewing the entire document, especially the figures, and effecting much contextual revision, including the above mentioned modifications. The manuscript was edited for publication by Susan K. Reed, Systems and Soft¬ ware Division. Action Summary The essential recommendations from this publication are summarized here to show the scope of these guidelines and to provide a quick overview of action items in establish¬ ing, implementing and maintaining a physical security program in an ADP facility. I. Organize The ADP Physical Security Program Assign responsibility for ADP Physical Security and establish a task force to prepare a plan for the ADP security program. Perform a preliminary risk analysis to identify major problem areas and select in¬ terim security measures as needed to correct major problem areas. II. Conduct A Risk Analysis Estimate potential losses to the ADP facility and its users from (1) physical destruc¬ tion or theft of physical assets; (2) loss or destruction of data and program files; (3) theft of information; (4) theft of indirect assets; and (5) delay or prevention of com¬ puter processing. Estimate the probability of occurrence for potential threats and their effect on the ADP facility in terms of the five classes of loss potential. Combine the estimates of loss potential and threat probability to develop an annual loss expectancy. Select the array of remedial measures which effects the greatest reduction in the an¬ nual loss expectancy at the least total cost. Remedial measures will include: (1) changes in the environment to reduce exposure; (2) measures to reduce the effect of a threat; (3) improved control procedures; (4) early detection; and (5) contingency plans. III. Determine Local Natural Disaster Probabilities Evaluate the fire safety of the ADP facility (building location, construction, occupancy and housekeeping) and provide required fire detection and extinguishment, and possibly a trained fire fighting brigade. Evaluate the exposure to flooding from internal and external sources. Where needed, provide flood protection for the building relocate ADP hardware, reroute plumbing lines and provide water damage/fiood-control equipment (pumps, tarpaulins, etc.) Evaluate resistance of the building to wind and water damage if exposed to hurricanes, tornadoes or other high winds. 5 [PDF page 8] FIPS PUB 31 IV. Initiate A Security Program Prepare a plan and a schedule for implementing selected remedial measures. Prepare and maintain a policy and plans handbook to include: (1) an ADP physical security policy statement; (2) mandatory security procedures; (3) security guidelines for system design, programming, testing, and maintenance; (4) contingency plans; (5) security indoctrination materials; and (6) a security audit program. V. Protect Supporting Utilities Estimate the number and duration of electric power transients, undervoltage condi¬ tions and power interruptions and their annual loss expectancy. Install appropriate protective equipment such as: voltage regulating transformers, dual power feeders, uninterruptible power supplies, on-site power generators and ADP power isolation circuits. Estimate annual loss expectancy from air conditioning failures considering required operation schedules, annual profiles of local temperature and humidity, and an esti¬ mated number and duration of air conditioning failures. Where necessary, increase reliability with redundant equipment, provide for emergency use of outside air and augment maintenance capability to decrease mean time to repair. Estimate the annual loss expectancy from teleprocessing circuit failures. Where cost is justified, increase reliability with redundant communications circuits and augment repair facilities to decrease the duration of interruptions. Software should be designed to min¬ imize the impact of errors caused by communications failures. Determine if ADP operations could be interrupted by the failure of other supporting utilities such as water, natural gas, steam, elevators or mail conveyors. If necessary, take steps to increase reliability and decrease the mean time to repair. VI. Optimize Computer Reliability Perform a failure analysis to estimate the number and duration of significant hard¬ ware failures and their impact on ADP operations. Estimate the annual loss expectancy from delays in performing urgent ADP tasks. Where cost is justified, increase system reliability by adding peripherals, multiple configurations, etc. Review maintenance fa¬ cilities. Record and analyze all hardware failures in order to identify failure trends promptly and optimize preventive maintenance. VII. Provide Physical Protection Identify critical ADP areas including the computer room, data control and conversion area, data file storage area, programmer’s area, forms storage area, maintenance area, and mechanical equipment room, and then provide adequate physical protection and access control. Protect against theft, vandalism, sabotage, espionage, civil disorder and other forced intrusions with improved lighting and intrusion detection systems, with physical bar¬ riers at doors, windows, and other openings, and with guards as required. Control access to critical areas and ADP facilities with conventional or electronic door locks; supervision by guards or receptionists over movement of people and materials; administrative procedures (sign-in logs, identification cards or badges, property passes and shipping/receiving forms) ; and other regulations. VIII. Add Internal Procedural Security Determine potential targets for fraud, theft or misuse of resources by analyzing the work How and the nature of ADP tasks performed. Incorporate procedures which will minimize exposure to loss. Such procedures may include (1) requiring cooperation be¬ tween two individuals to perform critical tasks; (2) performing additional checks and bounds comparisons; (3) formalizing standards for high risk operations; and (4) in¬ dependent quality control checks. 6 [PDF page 9] FIPS PUB 31 Designate critical positions in ADP management, system programming, program li¬ brary control, input/output control, exception processing, applications programming, data base management, quality control, internal audit and hardware maintenance and require appropriate pre-employment screening. Train and supervise all ADP personnel to assure understanding of, and compliance with, internal controls. Implement control and record keeping procedures for job initiation, scheduling and dis¬ tribution of output to prevent unauthorized processing. Control access to physical data files to assure that data integrity is maintained, storage media are protected, custody of data files is traceable and their unauthorized use is prevented. Manual and automatic audit trails should be utilized. Establish policy and procedures for program and data file retention to satisfy require¬ ments for (1) back-up operation; (2) compliance with applicable statutes and regula¬ tion; (3) audit and management review of operation; (4) statistical analysis of opera¬ tions; and (5) resolution of data integrity problems. Implement programming, testing and documentation standards which satisfy require¬ ments for (1) audit capability; (2) automated acceptance testing; (3) control pro¬ gram maintenance; (4) quality controls on input data; and (5) non-dependence on an individual’s knowledge of systems and programs. IX. Plan For Contingencies Compile a set of back-up plans which accommodate the expected range of emergency events requiring back-up operation. The objective of such contingency plans is to pro¬ tect users of the ADP facility against unacceptable loss. Document performance spe¬ cifications, operation instructions and technical requirements (system hardware and software, program and data files, and preprinted forms) for each emergency opera¬ tion Select and periodically use an emergency back-up off-site ADP facility. Participate in establishing their security program. Provide protection for the source documents, input and output data and programs while using the off-site facility and in transit. Establish procedures to assure that (1) current copies of needed back-up materials are retained at a secure off-site location; (2) adequate time is available from compat¬ ible off-site ADP facilities; and (3) back-up personnel will be available if needed. Plan for reconstruction of the ADP facility following destruction including specifica¬ tions of (1) floor space (quantity, live load rating, location, etc. by functional use) ; (2) partitions, electric power service, air conditioning, communications, security, fire safety, etc.; and (3) ADP hardware, office equipment and supplies. Coordinate ADP emergency plans for fire, flood, civil disorders, etc. with the Facility Self-Protection Plan to ensure life safety, limit damage, minimize disruption to ADP operations, and expedite repair. X. Develop Security Awareness Determine the security training requirements for the ADP staff, senior management, building staff, etc. Select and implement appropriate security awareness techniques such as (1) training lectures and seminars; (2) posters; (3) orientation booklets; (4) amendments to job descriptions making employees responsible for security; (5) publicity for local se¬ curity incidents, as well as others occurring at similar installations; and (6) rewards for employees who prevent breeches in security. Establish and publicize punitive measures. 7 [PDF page 10] FIPS PUB 31 XI. Audit Physical Security Establish an internal audit team with representatives from the agency’s audit, build¬ ing safety and security, ADP, and users’ organizations. Develop an audit plan and schedule which systematically validates all critical security and emergency measures. State in the audit report which measures require improvement or replacement. Use a check sheet (problem description, responsibility for action, action required and follow¬ up) for each major deficiency to assure prompt resolution. 1. ADP Security Analysis 1.0. Introduction The word security when applied to automatic data processing (ADP), is often taken to mean protection against wrongful disclosures or al¬ ternatively as protection against an aggressive attack on an ADP facility. However, Webster* defines secure as “. . . not likely to fail or give away; firm; strong; stable . . .”. These are cer¬ tainly desirable characteristics for an ADP fa¬ cility and they are included in the broader meaning of security that this handbook ad¬ dresses. It is intended to assist ADP managers and supporting agencies in defining specific ADP physical security requirements, developing and implementing sound physical security pro¬ grams, and establishing and conducting audits of these programs. Those who are users of ADP facilities can avail themselves of this handbook to evaluate the security of those facilities, to participate effectively in security planning and to plan for adequate back-up. A Federal ADP facility exists to support the accomplishment of the missions of its parent agency and other users. The objective of the physical security program is to see that all reasonable steps have been taken to prevent situations which would interfere with mission accomplishment, in other words, to operate an ADP facility that is “not likely to fail.” 1.1. Scope The scope of the handbook is defined in detail in section 1.2, but generally speaking, it is con¬ cerned with physical effects or situations which affect the ADP facility. Measures to achieve controlled accessibility, a term defined in the “Controlled Accessibility Bibliography” [46]* 1 as the use of technological measures of hard¬ * Webster’s New World Dictionary, 19.T7, The World Publish¬ ing Co., Cleveland. 1 Figures in brackets indicate literature references in Ap¬ pendix R at the end of this handbook. ware and software in a computer system to pro¬ tect data against unauthorized access, have been excluded from this handbook. Privacy and confidentiality are defined as concepts which have to do with the nature of the data and who is authorized to have access. It should be under¬ stood, however, that it is difficult to place rigid boundaries on the various aspects of ADP se¬ curity. A given measure will often achieve more than one objective. More than one discipline or function often will be required to deal effec¬ tively with a particular requirement, and so it is important to take a broad view of the sub¬ ject during the study and planning stages. The term ADP security planner is used here as a convenient title for the person (s) respon¬ sible for ADP security planning, but this should not be taken to mean that any one person can be expected to be competent in every area. In¬ deed, at each appropriate point sources of spe¬ cial knowledge are recommended. The manager of an ADP facility will derive the most from this handbok if he designates security as an on-going operational function, and provides ade¬ quate staff and budget to support the function. The procedure suggested here for developing and implementing a physical security program can be summarized as follows: • Analyze risk as the basis for development of a security policy. • Select and implement appropriate security measures to reduce exposure to losses. • Develop contingency plans for back-up op¬ eration, disaster recovery and emergencies. • Provide indoctrination and training for personnel. ® Plan and conduct continuing tests and au¬ dits and adjust security measures and con¬ tingency plans as needed. 8 [PDF page 11] FIPS PUB 31 1.2. Threats to ADP Operations This handbook deals with the threats to ADP property and capital equipment and the physi¬ cal hazards to continuing operation as outlined below: Unauthorized access by people to specific areas and equipment for the purpose of com¬ mitting acts such as theft, arson, vandalism, tampering, circumvention of internal controls, or improper physical access to information. These controls may include physical barriers such as fences or partitions, locked doors, re¬ ceptionists or guards at control points, elec¬ tronic devices such as closed circuit television and intrusion detectors, administrative pro¬ cedures such as restricted access, and special identification badges. Measures to minimize interruptions to data processing operations caused by ADP hard¬ ware failures. These measures may include introduction of redundancy in critical por¬ tions of the hardware configuration, preventive maintenance, and close monitoring and analysis of the causes of hardware failures. Failure of supporting utilities including elec¬ tric power, air conditioning, communications circuits, elevators or mail conveyors. Protec¬ tive measures may include redundancy of criti¬ cal elements, close monitoring or performance, physical protection against tampering or nat¬ ural disasters and provision of means for prompt repair. Natural disasters including floods, wind¬ storms, fires and earthquakes. Countermeas¬ ures include careful selection of the site for the ADP building, details of building design and construction and provision of means to protect against the effects of emergencies. Protection against human errors through ef¬ fective use of training, supervision and con¬ trols to minimize errors. Nonavailability of key personnel guarded against by cross-training for critical positions. Neighboring hazards such as close proximity to chemical or explosive operations, airports, high crime areas or the like. Protection may in¬ clude site selection, building design features, ex¬ clusion of such hazards from the ADP facility building and emergency planning. Tampering with input, programs, or data files for fraudulent purposes. In addition to physical access controls, internal controls and procedures (which may also protect against errors) are used to deter or detect such tam¬ pering. Compromise of data through interception of acoustical or electromagnetic emanations from ADP hardware. Countermeasures include iso¬ lation of ADP hardware from potential loca¬ tions of interception equipment, shielding of ADP hardware or the room in which it is lo¬ cated and filtering of power lines. (It is not within the purview of this handbook to deal with interceptions through wiretapping or other compromise of data communications circuits.) Of course, not every ADP facility will be faced with all of these threats. The impact of a given threat may depend on the geographic location of the ADP facility (earthquakes), the local environment (flooding), the potential value of property or data to a thief (blank check stock or information of value to a commodities speculator), or the perceived importance of the agency to activists and demonstrators or sub¬ versives. 1.3. Risk Analysis Experience has shown that a quantitative risk analysis will produce the following benefits: • Objectives of the security program are di¬ rectly related to the missions of the agency. • Those charged with selecting specific secu¬ rity measures have quantitative guidance on the amount of resources which it is reasonable to expend on each security measure. • Long range planners will have guidance in applying security considerations to such things as site selection, building design, hardware configurations and procurements, software systems and internal controls. • Criteria are generated for designing and evaluating contingency plans for back-up operation, recovery from disaster and deal¬ ing with emergencies. • An explicit security policy can be gen¬ erated which identifies what is to be pro¬ tected, which threats are significant and who shall be responsible for execution, re¬ view and reporting of the security pro¬ gram. For all these reasons, it is recommended that the ADP facility management begin develop¬ ment of the security program with a risk analy¬ sis. A suggested procedure is outlined in the sections which follow. 1.3.1 Loss Potential Estimate The first step of the risk analysis is to esti¬ mate the potential losses to which the ADP facility is exposed. The objective of the loss potential estimate is to identify critical aspects of the ADP facility operation and to place a dollar value on the loss estimate. Losses may result from a number of possible situations: Physical destruction or theft of tangible as¬ sets. The loss potential is the cost to replace lost assets and the cost to the user of delayed processing. 9 [PDF page 12] FIPS PUB 31 Loss of data or program files. The loss poten¬ tial is the cost to reconstruct the files either from back-up copies if available or from source documents and possibly the cost to the user of delayed processing. Theft of information. The loss potential here is difficult to quantify. Consider for example information gathered, collated and then pub¬ licly disseminated which affects marketplace activity. Knowledge of such information prior to dissemination would give a trader an ad¬ vantage over others who would in effect sus¬ tain a loss equal to the trader’s gain. Although the agency itself would sustain no direct loss it clearly would have failed in its mission. In some cases information itself may have market value as, for example, a proprietary software package or a name list which can be sold. Indirect theft of assets. If the ADP system is used to control other assets such as cash, items in inventory or authorization for per¬ formance of services, then it may also be used to steal such assets. The loss potential would be the value of such assets which might be stolen before the magnitude of the loss is large enough to assure detection. Delayed processing. Presumably every appli¬ cation has some time constraint on it and failure to complete it on time will cause a loss. In some cases it may be relatively easy to esti¬ mate the potential loss. For example, a failure to process payment checks promptly would prevent the exercise of a prompt payment discount under a procurement contract. Likewise, delays in an inventory system may lead to idle man¬ power at a warehouse, with secondary losses to recipients of materials stored at the ware¬ house, such as the cost of idle labor at a con¬ struction site. In other cases the loss potential may not be as obvious as, for example, a delay in issuing paychecks. Sometimes it may be helpful to use the daily operating cost of an agency as a rough rule-of-thumb estimate of the cost of delayed processing in those situa¬ tions where a delay would more or less halt operations of an agency. It should be noted that the loss experienced will in general increase with the duration of the delay. Therefore it is important to establish the maximum “no loss” delay time and an esti¬ mate of the median time to reconstruct the ADP facility after total destruction. Delay loss estimates, where losses are significant, should then be made for a range of delay durations between these two bounds. Generally three or four such representative durations will be ade¬ quate to establish loss trends. The estimate of physical destruction loss po¬ tential is quite straightforward. The ADP se¬ curity planner with the help of the building manager and procurement division should con¬ struct a table of replacement costs for physical assets of the ADP facility. This will usually include the following: The building itself. Special equipment installed to support the ADP facility such as air conditioning, elec¬ tric power distribution, raised floor. ADP hardware and other special equipment such as decollators, microfilm processors, keypunches. Supplies and materials such as magnetic tapes, disk packs, forms, ribbons. Office equipment such as desks, chairs, file cabinets, shelves, typewriters. Preparation of this tabulation, broken down by specific areas, with help to identify areas needing special attention. While the contents of the typical office area may be valued at $5 to $10 per square foot, it is not unusual to find that the contents of a computer room are worth $500 to $2000 per square foot. The estimate will also be helpful in planning for recovery in the event of a disaster as described in section 8.4. The remaining four loss potential types listed above are dependent on the characteristics of the individual data processing tasks performed by the ADP facility. The ADP security plan¬ ner should review each task to establish which losses it is exposed to and which factors affect the size of the potential loss. Undoubtedly, he will want to call on users to help make these estimates, since it is unlikely that he will be aware of all loss factors. In order to make the best use of time, the ADP security planner may want to do some kind of rapid, preliminary screening in order to identify the tasks which appear to have sig¬ nificant loss potential. For example, he might construct a table of preliminary estimates like the following very simple example: File Propri¬ Assets Task Run Recon- : Sensitive etary Con¬ No Cost Name Time struction Data Data trolled Delay P 1.5/D Easy No Yes Cash One day Q On line Very Diff. No No None 2 hours R 2.5/D Difficult Yes No Cash 8 hours S 2.0/W Uses P files No No None One week T 0.5/D Very Easy Yes No Inven- 4 days tory 10 [PDF page 13] FIPS PUB 31 In this example task P runs 1.5 hours per day, has files that are easy to reconstruct, has no sensitive data, but does have proprietary data, controls cash and appears on first inspection to be able to be delayed up to one day without significant cost. In actual practice, the ADP security planner would provide much more de¬ tail: what files are used and why they are easy or difficult to reconstruct, what data is pro¬ prietary and how much cash is processed. Having made the above analysis, he can then draw these initial conclusions: Loss Exposure Loss of Theft of Theft of Delayed Task Data Info. Assets Processing Q Yes No No Extreme R Yes Yes Yes Moderate P No Yes Yes Moderate T No Yes Yes Low S No No No Very Low Notice that on a judgmental basis, he has re¬ arranged the tasks in descending order of sen¬ sitivity. Tasks Q and R should probably receive early attention and detailed evaluation. Task S appears to have a low loss potential and prob¬ ably will require litle more than confirmation of the preliminary appraisal. Having made a preliminary screening to identify the critical tasks, the ADP security planner should seek to quantify their loss po¬ tential more precisely with the help of user representatives familiar with the critical tasks and their impact on other activities. He should think about what could go wrong and how losses could occur, under the assumption that if something can go wrong that it will. The fact that a given task has never been tam¬ pered with or used for an embezzlement is no assurance that it never will be. At this stage of the risk analysis, the ADP security planner should assume the worst. Later he will under¬ take to estimate probability of occurrence, but at this point he wants to identify all of the sig¬ nificant potential losses so that each of them will be addressed by the security program. 1.3.2. Threat Analysis The second step of the risk analysis is to evaluate the threats to the ADP facility. Threats and factors which influence their rela¬ tive importance have been outlined in section 1.2. Details of threats are given in the chapters which follow and, to the extent it is available, general information about the probability of oc¬ currence is given. These data and the applica¬ tion of common sense should be used by the ADP security planner to develop estimates of the probability of occurrence for each threat type. While the overall risk analysis should be conducted by the ADP security planner, others can contribute to the threat analysis and their help should be solicited. The following is a list of threats and suggested sources of help in analyzing them: Threat Sources of Refer to Information section Fire Building fire mar¬ shal and local fire department 2.1 Flood Army Corps of Engineers 2.2 Earthquake National Earth¬ quake Information Center 2.3 Windstorm National Oceanic and Atmospheric Administration and local National Weather Service Office 2.4 Power Failure Building engineer and local public utility 3.1 Air Condi¬ tioning Failure Building engineer and air condi¬ tioning vendor 3.2 Communica¬ tions Failure Federal Tele¬ communications System, building and local telephone company 3.3 ADP Hardware Failure Hardware vendors and Federal Supply Service 4.0 Intruders, Vandals, etc. Building manager, security director and the Office of Federal Protective Service Man¬ agement, GSA. 5.0 Compromising Emanations Hardware vendors and the Office of Federal Protective Service Man¬ agement, GSA. 5.2 Internal Theft or Misuse System Design, In¬ ternal Audit and Personnel Division 6.0 1.3.3. Annual Loss Expectancy The third step in the risk analysis is to com¬ bine the estimates of the value of potential loss and probability of loss to develop an esti¬ mate of annual loss expectancy. The purpose is to pinpoint the significant threats as a guide to the selection of security measures and to develop a yardstick for determining the amount of money which it is reasonable to spend on each of them. In other words, the cost of a given security measure should relate to the loss(es) against which it provides protection. 11 [PDF page 14] FIPS PUB 31 To develop the annual loss expectancy, one can construct a matrix of threats and potential losses. At each intersection one asks if the given threat could cause the given loss. For example, one might decide that fire, flood and sabotage do not cause theft-of-information losses but that in varying degrees all three result in physical destruction losses and losses due to delayed processing. Likewise internal tampering could cause an indirect theft of as¬ sets. In each case where there can be signifi¬ cant loss, one multiplies the loss potential by the probability of occurrence of the threat to generate an annual estimate of loss. As an example of a loss expectancy estimate, consider the simplified case where there are three ADP tasks in which loss could result from delays in completed processing as follows: Task Delay Duration One Hour Four Hours Eight Hours One Day A $10,000 $ 45,000 B $ 5.000 12,000 55.000 C $3,000 16,000 45,000 160,000 TOTAL $3,000 $21,000 $67,000 $260,000 Further assume that the annual probability of each such delay duration resulting from elec¬ tric power failures have been estimated to be 0.75, 0.31, 0.10 and 0.09 respectively. One could conclude that the annual loss expectancy from electric power failure would be: 0.75 x 83,000 + 0.31 x $21,000 + 0.10 x $67,000 + 0.09 X $260,000 = $38,860 per year. The cost of power failures is relatively easy to estimate since both probability of occurence and effect on operations can be quantified with some precision. Air conditioning and communi¬ cations failures also fall into this class. Quan¬ tifying fire losses is a different matter. One might deal with them by considering several degrees of severity and a number of loss types as shown in figure 1. The probabilities of oc¬ currence come from the estimate of inherent fire safety in section 2.1 and the dollar losses are from the estimates of loss potential in sec¬ tion 1.3.1. A similar technique can be applied to earthquakes, floods, windstorms and similar natural disasters. Fire Description Minor Fire in ADP Area Major Fire in Bldg, Total Loss Fire Occurrence Probability 0.10 0.05 .0005 Building Damage $10,000 $100,000 $3,700,000 ADP Hardware 50,000 10,000 2,100,000 ►4 General Equip. 5,000 285,000 73 a) a Supplies, etc. 10,000 _ 130,000 '£ >» fl H a> Task D—Delay 35,000 Task E—Delay 5,000 7,000 100,000 o Task F—Delay File Reconstruct 12,000 5,000 20,000 250,000 85,000 Total potential loss __— . 97,000 137,000 6,685,000 Annual loss __ - $ 9,700 $ 6,850 $ 3,342 Figure 1. Estimating fire loss. Human acts are more difficult to project since there is no easy way to estimate probability of occurrence. However, one can probably esti¬ mate potential losses with acceptable accuracy and so pinpoint critical threats. For example, consider fraud via program tampering. An ex¬ amination of tasks which disburse funds might reveal the following: Task Dollars per Cycle Expected Program Changes (next 12 months) J $20,000,000 5 K 200,000 25 L 5,000,000 10 If one assumed that a 1% theft would definitely be detected and also that the embezzler would not attempt to insert a wrongful program change more often than once in ten changes, one could draw these conclusions: Task Potential Theft Fraud Expectation Est. Loss J $200,000 0.5 $100,000 K 2,000 2.5 5,000 L 50,000 1.0 50,000 $155,000 12 Such conclusions might appear improbable. Perhaps the assumptions are not valid. The [PDF page 15] FIPS PUB 31 judgment factor plays a large part in arriving at these conclusions; repeated attempts may serve to sharpen one’s judgment in such mat¬ ters. As a result of iterative analyses, one might arrive at an annual loss for J of $10,000, or twice that of K, and for task L a loss equal to that for K or $5,000; the revised annual loss potential for the three tasks then would be only $20,000. The key point is that in attempting the esti¬ mate, a clearer picture of the critical exposures and reasonable criteria emerges. It now be¬ comes obvious that task K is just as critical as task J because, even though it disburses only one hundredth as much money per cycle, the program is still in a fluid state and therefore more subject to compromise. Because a quanti¬ tative effort has been undertaken, the proba¬ bility of occurrence of each threat and its effect on the ADP facility have been examined re¬ alistically. Clearly this is not an exact science. Indeed, it is quite likely that one will have to reappraise threats and losses more than once, concentrat¬ ing on the areas initially identified as most critical, before the loss expectancy estimate reaches a satisfactory level of confidence. In some cases it may not be feasible to generate more than a rough estimate; however, the value of disciplined thinking about risk will be ample reward for the effort to deal with it in a quantitative way. 1.3.4. Selecting Remedial Measures When the estimate of annual loss has been completed, ADP management will have a clear picture of the significant threats and critical ADP tasks. The response to significant threats can take one or more of the following forms: Alter the environment to reduce the proba¬ bility of occurrence. In an extreme case this could lead to relocation of the ADP facility to a less exposed location. Alternatively, a haz¬ ardous occupancy adjacent to or inside the ADP facility could be moved elsewhere. Erect barriers to ward off the threat. These might take the form of changes to strengthen the building against the effects of natural dis¬ asters, saboteurs or vandals. Special equipment can be installed to improve the quality and reliability of electric power. Special door locks, guards and intrusion detectors can be used to control access to critical areas. Improve procedures to close gaps in controls. These might include better controls over opera¬ tions, more rigorous pre-hire screening or standards for programming and software testing. Early detection of harmful situations permit more rapid response to minimize damage. Fire or intrusion detectors are both typical ex¬ amples. Contingency plans permit satisfactory accom¬ plishment of agency missions subsequent to a damaging event. Contingency plans will in¬ clude immediate response to emergencies to protect life and property and to limit damage, maintenance of plans and materials needed for back-up operation off-site and maintenance of plans for prompt recovery following major damage to or destruction of the ADP facility. The criteria for selecting specific remedial measures are that the annual cost of the re¬ medial measures shall be less than the reduc¬ tion in expected annual loss which they bring about and that the mix of remedial measures selected shall be the one having the lowest total cost. The first criterion simply says that there must be a cost justification for the security program—that it returns more in savings to the ADP facility than it costs. This may seem obvious but it is not uncommon for an ADP manager to call for a security measure without first analyzing the risks. His experience and judgment tell him that some particular action is desirable. While this might seem to obviate the need for risk analysis, what it really amounts to is recognition of a possibly serious but unquantified loss potential. It would be more appropriate for the ADP manager to factor his judgment into a quantified risk analysis. The second criterion reflects the fact that a given remedial measure may often be effective against more than one threat. To illustrate: REMEDIAL MEASURES THREATS Fire Internal theft External theft Hurri¬ cane Sabotage Fire detection system X X Loss control team X X X Roving guard patrol X X X X Intrusion detectors X X X Personnel screening X X On-site power generator X X Back-up plan X X X Since a given remedial measure may affect more than one threat, the least cost mix of measures probably will not be immediately ob¬ vious. One possible way to make the selection is to begin with the threat having the largest annual loss potential. Consider possible reme¬ dial measures and list those for which the an¬ nual cost is less than the expected reduction in annual loss. (Precision in estimating cost and loss reduction is not necessary at this point.) If two or more remedial measures would cause a loss reduction in the same area, list them all but note the redundancy. Repeat the process for the next most serious threat and continue until reaching the point where no cost justifi- 13 [PDF page 16] FIPS PUB 31 able measure for a threat can be found. When the cost of a remedial measure is increased if it is extended to cover an additional threat, the incremental cost should be noted. At this point one has a matrix of individual threats and remedial measures with estimates of loss re¬ ductions and costs and thus an estimate of the net saving, which can also be shown graphi¬ cally: THREATS REMEDIAL MEASURES A B C D J 20* 9 11 10 0 10 4 1 8 2 5 —3 K to o •X- 15 5 12 0 12 6 0 6 4 2 2 * Same effect. For each threat, the estimated loss reduc¬ tion, the cost of the remedial measure and the net loss reduction have been given (in that order). By applying remedial measure J to threat A at a cost of $9,000, a loss reduction of $20,000 can be expected (a net saving of $11,000). Furthermore remedial measure J will reduce the threat B loss by $10,000 at no addi¬ tional cost and the threat C loss by $4,000 at an added cost of only $1,000. Finally, though, it appears that it would cost more than it would save to apply J to threat D. Therefore J would not be implemented for D. The net loss reduc¬ tion from J could be expressed as: J (A, B & C) =11 + 10 + 3 = $24,000 The table indicates that J and K have the same reduction effect on threat A. Since K costs more than J, it might, at first glance, be re¬ jected. However, K (A, B, C & D) = 5 + 12 + 6 + 2 = $25,000 and J (A, B & C) + K (A, B, C & D) = — 4+22 + 9 + 2 = $29,000 Therefore, while J and K are equally effective on threat A, K appears to be more effective than J on the other threats, but further check¬ ing shows that their combined use results in the greatest overall net loss reduction. By going through the process just described, using preliminary estimates for cost and loss reduction, the ADP security planner can test various combinations of remedial measures. This will enable him to identify the subset of remedial measures which appears to be the most effective. At this point the ADP security plan¬ ner should review the estimates for the can¬ didate subset and refine them as necessary to establish confidence in the tentative choices. In marginal situations this might cause a change of the optimum subset. However, by iterating the process as required, the ADP security plan¬ ner will finally reach the point where he can recommend a given group of remedial meas¬ ures with considerable confidence. And, almost as important is the ability to defend the re¬ jection of remedial measures which cannot be cost justified. If all of the above procedures have been followed, the following will have been estab¬ lished and documented: ® Significant threats and probabilities of oc¬ currence. • Critical tasks and the loss of potential re¬ lated to each threat on an annual basis. • A list of remedial measures which will yield the greatest net reduction in losses, together with their annual cost. With this information at hand ADP manage¬ ment can move ahead with implementation of the physical security program. Since the analy¬ sis of remedial measures will have identified those with the greatest impact, relative priori¬ ties for implementation can also be established. 1.4. Implementing the Security Program In section 1.3 the use of a risk analysis has been described as the basis for developing an ADP security program. Implementation of the program will depend on local conditions and the practical constraints of time and budget, but it may not always be clear just where to begin. The following is a brief outline of a pro¬ cedure which should be generally applicable. • Preliminary planning. Establish an ADP security study team to prepare an ADP secu¬ rity program consisting of detailed task de¬ scriptions for the next three tasks, a budget and schedule and responsibility assignments. • Perform a preliminary risk analysis to identify major problem areas. • Select and implement urgent “quick fix” security measures as needed. • Perform and document a detailed risk analysis for review and approval • Based on the approved risk analysis se¬ lected, cost justify and document action plans with budgets and schedules for security meas¬ ures, contingency plans, training and indoctri¬ nation plans and test and audit plans. 14 [PDF page 17] FIPS PUB 31 • Carry out the approved action plans. • Depending on the results of tests, audits and changes in mission or environment, repeat the detailed risk analysis and subsequent steps on a regular, at least annual, basis. The action plans should include adequate documentation. The documentation might in¬ clude : • A security policy statement which provides general guidance and assigns responsibilities. • A security handbook which describes in detail the security program and procedures and the obligations of ADP personnel, users and supporting personnel. • Technical standards for system design, programming, testing and maintenance to re¬ flect security objectives. • Contingency plans for back-up operations, disaster recovery and emergency response. • Booklets for ADP staff indoctrination in security program requirements. Depending on the normal practice of the ADP facility, these documents may be com¬ pletely separate items or may be included in other documents. For example, emergency re¬ sponse plans for the ADP facility might be included in the agency’s Facility Self-Protec¬ tion Plan. Similarly, technical security stand¬ ards could be added to existing documents. The final point to be made is the importance of continuing audit and review of the security program. A major effort will be required for the initial risk analysis but once it has been com¬ pleted a regular review and updating can be done much more quickly. By evaluating changes in agency mission, the local environment, the hardware configuration and tasks performed, the ADP security planner can determine what, if any, changes should be made in the security program to keep it effective. 1.5. Supporting Documents There are a number of Federal documents relating generally to ADP security which will be helpful to security planners. These, as well as a number of other useful references, are listed in the bibliography in Appendix B. It is suggested that this list be consulted by plan¬ ners early in their assignment in order to be able to take advantage of the extensive fund of knowledge they represent. 2. Anticipating Natural Disasters 2.0. Introduction This chapter deals with fire, flood, windstorm and earthquake. These events all tend to have the same basic effects on ADP operations: physical destruction of the facility and its con¬ tents and interruption of normal operations. They also represent a threat to the life safety of the ADP staff. In the sections which follow, protective measures and factors for evaluating exposure are presented. Planning for emergency response is discussed in Chapter 8—Contin¬ gency Planning. 2.1. Fire Safety Experience over the last two decades has demonstrated the sensitivity of ADP facilities to fire damage and disruption of operations. For example, a parts warehouse which included a $1 million computer system was totally de¬ stroyed by a fire. The building, almost 0.8 sq mi (two hectares) in size, was of non-com¬ bustible construction and had neither sprin¬ klers, interior fire partitions nor fire curtains. Furthermore, the building was located just out¬ side the municipal fire district, presumably be¬ cause of the low tax rate. The fire evidently started when an electric spark ignited a flam¬ mable solvent being used to remove floor seal¬ ant. Although the structure, contents and com¬ puter system were completely destroyed, the company’s emergency procedures called for storage of magnetic tapes in fire-rated vaults and they were recovered intact. As a result, and with a major effort on the part of the hard¬ ware vendor, a new computer system was op¬ erating at an alternate site four days later. This episode highlights the value of close at¬ tention to both fire safety and contingency planning. A number of such major losses have involved noncombustible buildings. In those cases where vital tapes had been safeguarded and the computer hardware was relatively un¬ complicated, rapid recovery was possible, often in a matter of days. However, it seems likely that if a large computer configuration is de¬ stroyed or if back-up records are inadequate, recovery would be a lengthy process that could take many weeks or months. Fire safety should be a key part of the ADP facility physical security program and should include these elements: 15 [PDF page 18] FIPS PUB 31 • Location, design, construction and main¬ tenance of the ADP facility to minimize the exposure to fire damage. • Measures to insure prompt detection of and response to a fire emergency. ® Provision of adequate means to extin¬ guish fires and for quick human interven¬ tion. • Provision of adequate means and personnel to limit damage and effect prompt re¬ covery. Each of these points is discussed in the sub¬ sections which follow. A comprehensive treat¬ ment of the subject of fire prevention and con¬ trol is also the subject of the Fire Protection Handbook [24], To quote from the handbook itself, it “. . . constitutes an authoritative en¬ cyclopedia on fire and its control and is designed to serve both as a textbook for those learning the science and as an independent reference book . . .”. It includes fire control considera¬ tions in building design and construction, tables of the fire hazard properties of several hundred materials, and an engineering handbook on hydraulic properties, in addition to the other topics on fire control one would expect in such a handbook. 2.1.1. ADP Facility Fire Exposure The first factor to consider in evaluating the fire safety of an ADP facility is what fire expo¬ sure results from the nature of the occupancy of nearby buildings and the ADP facility building. Generally speaking the degree of hazard as¬ sociated with a given occupancy depends on the amount of combustible materials, the ease with which they can be ignited and the likeli¬ hood of a source of ignition. The following oc¬ cupancies have been found to be particularly hazardous: building under construction; cloth¬ ing and textile processing; chemical, plastic, paint and petroleum processing; electric appli¬ ance assembly; foundries; paper manufactur¬ ing; and storage and warehousing operations. The inherent hazard of an occupancy can also be evaluated in terms of the probable severity of a fire as a function of the heat potential (fuel load) of the contents. This relationship can be expressed approximately as follows: Fuel Loading (Equivalent pounds of wood per square foot) Potential Heat Release Kilo¬ calories per square centimeter) Fire Severity (duration in hours) 5 11 0.5 10 22 1 20 43 2 30 65 3 50 110 6 70 152 9 A typical office with metal furniture and storage cabinets will have fuel loading rang¬ ing from 5 to 15 pounds per square foot (11 to 33 kcal/cm2)- A storage room for paper forms and boxed punched cards, or a magnetic tape library, will have fuel loads of 50 to 80 pounds per square foot (110 to 175 kcal/cm2).* The severity of a fire and its effect on the structure and contents will depend on the rate at which temperature rises and the duration of the fire. Thus if the fuel load is so configured and stored as to retard ignition and combustion of, for ex¬ ample, paper records in metal file cabinets, temperature will rise relatively slowly. If the same fuel load were in the form of reels of magnetic tape stored in relatively open racks, the temperature could be expected to rise rap¬ idly but the fire would be of brief duration. The second fire safety factor is the design and construction of the building. There are five basic types of construction: • Fire-Resistive—The structure of the build¬ ing—framing, floors, walls and roof—is constructed of noncombustible materials which are insulated to protect against loss of strength as a result of a fire. • Heavy Timber—Exterior walls are non¬ combustible with a 2-hour rating and col¬ umns, beams, floors and roof are heavy timber. Because of the slow burning char¬ acter of heavy timber, it will be superior in performance to noncombustible. • Noncombustible—The structure is non¬ combustible, but lacks protection against the effect of heat on the structural mem¬ bers. The difference is this: while a fire in a noncombustible building will not draw fuel from the structure itself, the heat from the fire may cause the structure to collapse. A classic example of a noncom¬ bustible building fire was a transmission plant in Michigan. Although the structure itself did not contribute any fuel to the fire, the asphalt on the roof provided enough fuel to completely destroy the building. • Ordinary Construction — Ordinary con¬ struction is the same as Heavy Timber ex¬ cept that the dimensions of the timber portions of the structure are too small to qualify as heavy timber. * NFPA computes fuel load based on a heat of combustion of 8,000 RTU per pound : a representative value for wood or paper. Magnetic tape is roughly twice as combustible as w'OOd, so that 40 lb of magnetic tape would have an 80 lb fuel load. 16 [PDF page 19] FIPS PUB 31 • Wood Frame—This is the typical residen¬ tial construction using two inch (5 cm) thick framing and one inch (2.5 cm) boards. To summarize the above simply, and ignor¬ ing design features which can increase fire re¬ sistance, one can construct the following table: Type of Construction Approximate Fire Classification Fire Resistant 2 or 3 hours Heavy Timber 1 plus hours Noneombustible 1 hour Ordinary Construction Less than 1 hour Wood Frame Minutes The actual performance of a building will depend not only on the type of construction, but on design details such as: • Fire walls which in effect divide a struc¬ ture into separate buildings with respect to fires. • Fire rated partitions which retard the spread of a fire within a building. • Fire rated stairwells, dampers or shutters in ducts, fire stops at the junction of floors and walls and similar measures to retard the spread of smoke and fire within a building. • Use of low-flame-spread materials for floor, wall and ceiling finish to retard propaga¬ tion of flame. To summarize, the four basic fire safety fac¬ tors and their effects can be tabulated as follows: Factor Effect Occupancy Fuel load Construction Type Construction Details Probability of a fire occurring Intensity and duration of a fire Resistance to structure damage Confinement of a fire It should be understood that this discussion has been much simplified. However, considera¬ tion of these factors by the ADP security plan¬ ner as they apply to an existing or projected ADP facility will help him to determine the amount of attention he should pay to fire safety. He will want to seek the assistance of a quali¬ fied fire protection engineer in evaluating the inherent fire safety of the ADP facility and identifying hazards. A detailed discussion will be found in “Building Firesafety Criteria” [13]. The July, 1973 fire at the U.S. Military Per¬ sonnel Records Center, Overland, Mo., was an unfortunate demonstration of the result when well tested fire safety design criteria are dis¬ regarded in overemphasizing protection against other risks. Lack of sprinkler protection, in¬ adequate access to the fire site and related de¬ sign deficiencies seriously hampered fire fight¬ ing and in the end resulted in much more damage to records than would have resulted from the operation of sprinkler heads. The third factor in fire safety is the way in which the building is operated. It should be understood that the inherent fire safety of a building can be rendered ineffective by careless operation. This includes: fire doors propped open; undue accumulation of debris or trash; careless use of flammable fluids, welding equip¬ ment and cutting torches; substandard electric wiring; inadequate maintenance of safety con¬ trols on ovens and boilers; and excessive con¬ centration of flammable materials. ADP facili¬ ties, for example, have a particular hazard from the accumulation of lint from card and paper operations. The ADP physical security program should strive, in coordination with the building maintenance staff, to identify and eliminate such dangerous conditions. Furthermore, it should be understood that this must be a con¬ tinuing effort and a consideration in the assign¬ ment of security management responsibilities. The security audit plan should include verifica¬ tion of compliance with established standards. Specific guidance for the construction of ADP facilities will be found in chapter 2 of “Fire Protection for Essential Electronic Equip¬ ment” [9], This document, hereafter referred to as RP-1, has been adopted by the GSA for all GSA facilities under GSA Order PBS 5920.4B with certain minor modifications. 2.1.2. Fire Detection Despite careful attention to the location, de¬ sign, construction and operation of the ADP facility, there is still the possibility that a fire can start. Experience has shown repeatedly that prompt detection is a major factor in limiting fire damage. Typically a fire goes through three stages. Some event, such as a failure of electrical insulation, causes ignition. An electrical fire will often smolder for a long period of time. When an open flame develops, the fire spreads through direct flame contact, progressing relatively slowly, with a rise in the temperature of the surrounding air. The dura¬ tion of this stage is dependent on the com¬ bustibility of the materials at and near the point of ignition. Finally the temperature reaches the point at which adjacent combusti¬ ble materials give off flammable gases. At this point the fire spreads rapidly and ignition of nearby materials will result from heat radia- 17 [PDF page 20] FIPS PUB 31 tion as well as direct flame contact. Because of the high temperatures and volumes of smoke and toxic gases associated with this third stage, fire fighting becomes increasingly difficult and often people cannot remain at the fire site. Given the objective to discover and deal with a fire before it reaches the third stage, one can see the limitation of fire detection which depends on detecting a rise in air temperature. It is for this reason that RP-1 requires that the areas in which electronic equipment is in¬ stalled be equipped with products-of-combus- tion (smoke) detectors. Such detectors use electronic circuitry to detect the presence of abnormal constituents in the air which are usually associated with combustion. To be effective in providing prompt detection the following points should be considered in designing a fire detection system: • The location and spacing of detectors should take into consideration the direc¬ tion and velocity of air flow, the presence of areas with stagnant air, and the loca¬ tion of equipment and other potential fire sites. Note that detectors may be required under the raised floor, above the hung ceiling and in air conditioning ducts as well as at the ceiling. It may also be wise to put detectors in electric and telephone equipment closets and cable tunnels. • The design of the detection control panel should make it easy to identify the de¬ tector which has alarmed. This implies that the detectors in definable areas (for example, the tape vault, the east end of the computer room, etc.) should be dis¬ played as a group on the control panel. In other words, when an alarm sounds, in¬ spection of the control panel should indi¬ cate which area or zone caused the alarm. Generally, and preferably, each detector will include a pilot light which lights when the detector is in the alarm state. In some cases it may be determined that there should be a separate indicator light at the control panel for each detector. It is also important to see that the alarm system it¬ self is secure. Its design should cause a trouble alarm to sound if any portion of it fails, or if there is a power failure. Steps should be taken to assure that the system could not be deactivated readily, either maliciously or accidentally. In a recent case of suspected arson in a tape library it ap¬ peared that the smoke detection system had been turned off. • Meaningful human response to the detec¬ tion and alarm systems is necessary if they are to be of any value. This means that the fire detection system should be designed to assure that someone will al¬ ways be alerted to the fire. Typically, we expect that the computer room staff will respond to an alarm from the ADP facility alarm system. A remote alarm should also be located at another point in the building which we expect will be manned at all times, such as the lobby guard post, se¬ curity center or building engineer’s sta¬ tion. This provides for back-up response and response when the computer area is not occupied. If there is any possibility that the remote alarm point will not be manned at all times, a third alarm point should be located off-site, typically at the nearest fire station or location of the fire brigade for the facility. ® Proper maintenance is essential to the fire detection system. The nature of smoke de¬ tectors is such that nuisance alarms may be caused by dust in the air or other fac¬ tors. Thus there is a tendency to reduce sensitivity in order to eliminate nuisance alarms, with the result that detection of an actual fire may be delayed. To insure proper operation, it is important to see that qualified personnel (a vendor repre¬ sentative or building engineer) verify cor¬ rect operation at the time of installation and at least once each year thereafter. Furthermore, each fault condition should be corrected immediately. Unfortunately, there is a common tendency to turn off the fire detection system or silence the alarm bell, creating the danger that there will be no response if a fire should occur. In addition to alerting personnel to the pres¬ ence of a fire, the detection equipment can be used to control the air conditioning system. There is some support for the view that upon detection, air handling equipment should be shut down automatically to avoid “fanning the flames” and spreading smoke. This may not be the best plan, as nuisance alarms will result in needless disruption. A preferred technique may be to cause the system to exhaust smoke by stopping recirculation and switching to 100% outside air intake and room air discharge. As a rule this can be done by adjustment of air conditioning damper controls and their inter¬ connection with the fire detection system. How¬ ever, it may be necessary to modify the air conditioning system. More details will be found in section 3.2. 2.1.3. Fire Extinguishment Fire extinguishment is accomplished in four ways: • portable or hand extinguishers operated by agency personnel in an effort to control the fire before it gets out of hand. 18 [PDF page 21] FIPS PUB 31 ♦ • hose lines used by professional fire fighters to attack the fire with water. • automatic sprinkler systems which release water from one or more sprinkler heads when the air temperature reaches the de¬ sign temperature of the head which range from 135-280 °F (57-138 °C). • volume extinguishment systems using HA¬ LON-1301* which fill the room with a gas that interferes with the combustion proc¬ ess. A review of the history of fires involving electronic equipment and the effectiveness of each of these extinguishment devices has led the Federal Fire Council to establish a number of requirements for extinguishment in Chap¬ ter 3 of RP-1. First, at least one carbon dioxide extin¬ guisher of 15 pounds (6.8 kg) capacity or more and one 2(A gallon (9.5 1.) plain water ex¬ tinguisher shall be located within fifty feet (15 m) of each piece of equipment. These extin¬ guishers are intended to be used by ADP fa¬ cility personnel for immediate fire fighting. Given prompt detection and response by prop¬ erly trained personnel and freedom from gross fire hazards in the computer area, portable extinguishers will be effective for controlling most fires quickly. To insure effectiveness of portable extin¬ guishers, several points must be considered. Extinguishers should be placed in readily ac¬ cessible locations, not in corners or behind equipment. Each location should be marked for rapid identification; for example, a large red spot or band can be painted on the wall or around the column above the point where each extinguisher is mounted. It is important to have all extinguishers inspected. (See “Port¬ able Fire Extinguishers” [44].) Each extin¬ guisher should have an inspection tag affixed to it on which the inspector signs his name and gives the inspection date. In addition to the required extinguishers, it may be wise to provide five pound (2.3 kg) carbon-dioxide ex¬ tinguishers in areas principally staffed by per¬ sonnel unable to lift heavy objects. Experience indicates that such personnel can deal effec¬ tively with minor equipment and trash fires if lighter extinguishers are made available to them. The second RP-1 requirement is that com¬ puter areas be equipped with automatic sprink¬ lers and, unless building construction is fire resistive or noncombustible, that the entire building shall be so equipped. Portions of the building which are not protected by sprinklers * HALON—1301 is a term applied to bromotrifluoromethane, n lialogenated extinguishing agent. and which cannot be reached easily with hose lines from the exterior should have standpipes and inside hose systems. The automatic sprink¬ ler system is the preferred extinguishment sys¬ tem for a number of reasons, but the ADP facility manager may be concerned that in¬ stallation of sprinklers will expose the ADP facility to serious water damage. If the worst thing that could happen to an ADP facility were to spray water on the hardware, it would make sense to omit sprinkler protection, but it isn’t; the worst is a structural collapse of the building. In an effort to provide effective ex¬ tinguishment without damaging side effects, one might consider a IIALON-1301 deluge sys¬ tem. Carbon dioxide (C02) systems represent a significant life safety hazard and their use cannot be recommended. The characteristics of automatic sprinklers and HALON-1301 are compared below: Automatic Sprinklers HALON- 1301 Extinguishment mechanism Water cooling and smothering of fire site. Chemical inter¬ ference with com¬ bustion process. Reliability Very high ; limited by reliability of water supply. Very high ; limited by reliability of detection system. Effectiveness Very high. Very high if effec¬ tive concentration is achieved at fire site. Life safety hazard None. Some danger if concentration greater than 10%. Side effects Prompt cooling and cleaning of air by water spray with attendant damage to contents. No side effects if effective; otherwise corrosive toxic de¬ composition products. Approx, cost to install $1.00/sq. ft. new building, $3.00+/ sq. ft., retrofit. •$0.50/cu. ft. of pro¬ tected volume. Discharge controlled by: Air temperature (or auto, recycle) Detection system or manual. Time and cost to refurbish after fire Minutes and $5 to $20. Hours and 40% of installed cost. Because of its lower cost, proven effectiveness and inherent safety, the automatic sprinkler is the preferred fixed extinguishment system in most cases. IIALON-1301 appears to be better suited for the initial fire attack at critical points, such as a tape or disk storage area or a room housing one-of-a-kind hardware or at points which cannot be covered effectively by a sprinkler system, e.g., under a raised floor or in a cable tunnel. t 19 [PDF page 22] FIPS PUB 31 Automatic sprinkler systems offer a feature which should be included in the fire safety sys¬ tem. Devices called flow sensors are available which can be inserted into the sprinkler pipes to detect the flow of water. These flow alarms should be located at the source of water and at each major branch in the piping and should be connected to a fire alarm panel. When a fire causes a sprinkler head to open and discharge water, an alarm will be sounded alerting per¬ sonnel to the emergency. This feature can be of real value during hours when work areas are unoccupied, as the security force is alerted immediately to sprinkler operation and can shut off the water flow as soon as the fire is extinguished. To make this easy to do, the sprinkler system piping should be configured to supply the computer area from a single point and equipped with a shut-off valve which is located in an easily accessible point. All sprinkler system shut-off valves should have supervisory switches attached which will sig¬ nal the fire alarm panel if a valve is closed. This is important because there have been many cases where fires were not defeated be¬ cause sprinkler control valves had been left closed inadvertently. In some cases it was sus¬ pected that valves were closed deliberately. The gas extinguishment systems also have features which contribute to more effective and reliable quenching. Pressure sensors are used to detect a significant loss of gas and to signal a trouble alarm. Systems are normally installed so that there is a delay of up to a minute be¬ tween the initial alarm and release of the gas. With carbon dioxide systems, this allows the area to be clear of personnel, because of the serious hazard to life when the gas is dis¬ charged. With HALON systems, the delay per¬ mits the actual discharge of this rather ex¬ pensive quenching agent to be overridden manually when there is no fire or when the fire is quenched easily by using portable extin¬ guishers. If fire extinguishing equipment is to remain effective, it must have regular maintenance by properly qualified personnel. “Fire Extinguish¬ ing Equipment” [11] is a useful guide to ex¬ tinguisher equipment inspection and mainte¬ nance. The ADP security planner should work with the Building Manager and Fire Marshal to insure that an effective maintenance pro¬ gram is in effect. The bibliography lists a num¬ ber of standards, guidelines and recommenda¬ tions from the National Fire Code published by the National Fire Protection Association [22- 43]. 2.1.4 Fire Fighting The discussion of extinguishment has stressed the value of prompt, effective fire fighting. With regard to who should do this fire fighting, the ADP facility manager should consider local conditions carefully to determine the most practical approach to meet this in¬ dividual problem. Some ADP facilities are lo¬ cated within large industrial complexes which either employ their own professional firefight¬ ers, have highly trained industrial fire brigades or are located in close proximity to a municipal fire department composed of professional fire¬ fighters. Conversely, some facilities may be situated in remote locations where response by professional or highly trained firefighters is de¬ layed or perhaps nonexistent because of travel distance. Obviously, the best arrangement is one which results in immediate response by professional firefighters in time of need. How¬ ever, when this is not feasible, other alterna¬ tives must be explored—particularly when one considers the high value of equipment usually housed within ADP facilities. In all probability, the enlightened ADP fa¬ cility manager will want to establish a first line of defense against fire involvement between the time of notification of and response by pro¬ fessional or highly trained firefighters, and will incorporate this as part of the Facility Self Protection Plan. Every plant, regardless of size, needs personnel who are knowledgeable and trained in fire safety. Any practical and effective organization for fire protection must be designed to assure prompt action immedi¬ ately at the point where a fire breaks out. This usually necessitates every organizational unit or area of a plant having a nucleus of key em¬ ployees who are prepared through instruction and training to extinguish fires promptly in their incipient stage. Such individuals become knowledgeable in specialized fire protection and the systems applicable to the facility in ques¬ tion: how to turn in an alarm, which type of extinguisher to use for which type of fire and how to use it. Further, such individuals can serve as on-the-job fire inspectors, constantly seeking out and reporting and correcting con¬ ditions that may cause fires. They can help en¬ sure that fire fighting equipment is properly located and maintained, that storage does not cause congestion which could hamper fire fight¬ ing, and that general housekeeping is main¬ tained at a reasonably high level to minimize fire risk. Should a decision be made to establish an ADP facility fire brigade organization, refer¬ ence should be made to the NFPA “Industrial Fire Brigades Training Manual” [27]. This document will serve as a useful guide in or¬ ganizing and training a fire brigade. The ADP fire brigade should consist of a fire captain, a deputy fire captain and several fire fighters on each operating shift. Large ADP facilities should consider more fire fighters to ensure adequate coverage. All other members of the facility staff should vacate the premises during fire involvement. 20 [PDF page 23] FIPS PUB 31 Designated fire fighters should receive train¬ ing each year in extinguishing actual fires using extinguishers of the type located in the computer area. In addition, they should under¬ stand the operation of fire detection equipment, alarms, sprinklers and any other fire safety equipment. To maintain competence, the fire brigade should meet regularly, perhaps at two or three month intervals, for brief training sessions. The fire captain should review any new equipment or procedures. He might also lead a discussion about how to deal with a hypothetical fire situation with questions like: What equipment should be turned off? Where is the nearest extinguisher? Other nearby ex¬ tinguishers? Would there be any difficulty in getting at the fire site? Who is notified and how? He should also ask for discussion of newly-observed fire safety problems. Undoubt¬ edly the building fire marshal and the local fire department can and will contribute to the training program with training materials and facilities and with advice. Because of the special characteristics of ADP hardware and the desire to avoid disruption to operations, it is important for fire fighting and loss control measures to be carefully struc¬ tured. ADP management and systems and op¬ erations supervisors should participate with the fire marshal and fire captains in develop¬ ing guidelines for decisions to power down hardware, shut off air conditioning and take related steps. All fire control measures must be coordinated with the fire department serving the ADP installation. There should be site visits to familiarize the fire department with normal and emergency entrances, electric power switches, hoses and portable extinguish¬ ers, sprinkler control valves, location of covers for equipment, exhaust fans and ventilation controls, combustibles storage, building con¬ struction and characteristics, and other perti¬ nent items. Unique ADP hazards such as the susceptibility of disk and drum surfaces to contamination and the presence of underfloor electric outlets should be pointed out. Emergency planning is presented in more detail in Chapter 8. 2.2. Flood The discussion of automatic sprinklers in the preceding section may have left the impression that water damage can be dismissed as a sig¬ nificant threat to ADP facilities. While it is true that the damage resulting from operation of one or two sprinkler heads will be minor and certainly preferable to the smoke and heat damage of a major fire, flooding is quite a different matter. The water may be contami¬ nated with dirt, oil or chemicals. Buildings may be damaged or even destroyed. Tropical storm Agnes which swept through Pennsylvania in June, 1972, caused severe flooding. Newspaper accounts reported that hundreds of computer systems were submerged in mud and water. The resulting damage ap¬ peared to depend largely on location and the reported time to recover ranged from two days to two months. The Pennsylvania Bureau of Management Information Systems reported its large computer submerged in six feet of water. The entire reserve supply of certain forms used weekly, 45 million in all, was lost by another computer facility, leaving only a one week supply on hand. A number of computer centers lost card data files which were not backed up. This experience points up two things. First, if an ADP facility is located in a basement in a low lying area, disruptions from flooding are almost inevitable. Second, careful planning for back-up operation can greatly reduce the time required to restore normal operations after an emergency. Executive Order 11296 was issued in August 1966 in response to growing concern about flood-related losses in Federal buildings but to insure optimum use of flood plains by Federal agencies. In summary this Executive Order requires executive agencies to evaluate flood hazards when locating new facilities, adminis¬ tering funds to support facilities, evaluating future use of Federal facilities to be disposed of, or when planning land use so as to “pre¬ clude the uneconomic, hazardous, or unneces¬ sary use of flood plains . . .”. Where practical and economically feasible, it requires that flood-proofing measures be applied to existing structures. Flood hazard information is available pri¬ marily from the Army Corps of Engineers, the Tennessee Valley Authority and also from the Departments of Agriculture, Interior, Com¬ merce, Housing and Urban Development and from the Office of Emergency Planning. State and local agencies may also have information available about past floods. Basic guidelines are presented in “Flood Hazard Evaluation Guidelines for Federal Executive Agencies” [54]. These guidelines point out that there are three types of flood areas where flooding can be hazardous. First are riverine flood plains where floods are due to heavy rainfall or snow¬ melt runoff or to obstruction of a narrow chan¬ nel. Second are coastal flood plains bordering on a body of standing water where floods can result from high tides, wind-driven waves, tsunamis (large waves caused by undersea earthquakes) or from a combination of these effects. Finally, debris cones, deposited at the base of a mountain by mountain streams, are subject to flash flooding. If it appears that the ADP facility is located in any of these areas, one must give consideration to flood exposure. ♦ 21 [PDF page 24] FIPS PUB 31 In evaluating the exposure to natural flood¬ ing, the ADP security planner should first examine the rules and regulations issued by his agency under Executive Order 11296. Next he should examine such evaluations of flood haz¬ ard as may be available for his own building or other nearby Federal buildings. These should help to determine the need to look more closely at the exposure. The information avail¬ able will often allow the ADP security planner to estimate the probability of flooding to sev¬ eral levels. By examining the building layout, he can then estimate the probable effect on operations from damage or destruction of con¬ tents, interruption of electric power and com¬ munications, lack of access to the building, and the like. By relating these effects to the risk analysis, he will be able to estimate flood- related losses as a basis for cost justification of flood protection measures. In addition to the overall effect of natural flooding, one should examine the flood damage potential from all causes. The first step is to evaluate the location of the ADP facility within the building. The basement is potentially the least desirable location since surface water from heavy rain or fire fighting water may collect in the basement. Drains can be equipped with backwater or check valves to prevent back up. Electrically driven sump pumps and ejector pumps may be provided to augment gravity drainage. However, in an emergency situation these may all prove ineffective. Dur¬ ing a fire on an upper floor, the pumps and drains may be overwhelmed since fire fighting hose streams can easily pump a thousand or more gallons of water per minute into the building. Furthermore, it is possible that debris from the fire area may clog drains and pumps. Electric power for sump pump motors may be interrupted by a fire or hurricane—putting them out of service just when they are most needed. The ADP security planner should at¬ tempt to balance the physical protection offered by a basement location against the exposure to flooding and make a judgment about the net exposure. If the ADP facility is located in the basement and the flooding exposure is signifi¬ cant, it may be prudent to consider these countermeasures: • Sump pumps (one or more) driven by gas¬ oline motors for emergency use. • Drains equipped with check valves. • If surface water flooding is a significant threat, a supply of sandbags can be kept on hand to be used to construct a dike quickly. Heavy duty adhesive tape may be adequate to seal low lying exterior doors. • It may be possible to install masonry curbs around the ADP area to divert flood water. This will help only with minor flooding but may be worth the effort. These measures will be helpful where the exposure is modest or comes primarily from internal sources. For existing facilities having a significant exposure to external flooding, full scale flood proofing may be required. Excellent guidance will be found in “Flood-Proofing Reg¬ ulations” [51]. This document is in the form of a model building code and provides guidance for minimizing flood-related hazards of build¬ ing occupancy and for protecting structures against flood damage. Flooding may also result from plumbing leaks. As a part of the threat evaluation, the ceiling above the ADP facility should be in¬ spected for plumbing lines and for holes. Ideally no pipes should be routed over ADP hardware areas; where this is unavoidable, easily accessible shut-off valves should be pro¬ vided. Likewise, chilled or condenser water pipes which suport air conditioning units in¬ side the ADP area should have shut-off valves which can be used to isolate a leak. Major water lines should be instrumented to detect abrupt loss of pressure—a sign of catastrophic failure—to alert the building engineer and, perhaps, shut off pumps automatically so as to limit the amount of water which can escape. All holes in the floor slab over the ADP facility should be plugged with cement or similar ma¬ terial. Many buildings include so called wet columns. These are structural columns with adjacent vertical plumbing lines usually re¬ ferred to as risers. As a rule one can identify a wet column because the walls enclosing it will be larger than most columns to allow space for the pipes. Since wet columns represent an increased exposure to leaks or flooding it would be preferable to exclude them from ADP areas. When this is unavoidable, each column should be checked to insure that any leakage will drain freely to the floor below. Almost all computer rooms are equipped with a raised floor to provide a protected space for inter-cabinet and power cables (and often as a supply air plenum for the air conditioning system). If water collects under the raised floor, there is a danger that these cables will be affected. Inter-cabinet cables with connectors at the ends only should be highly water resist¬ ant. However, power cables often plug into receptacles located on the floor, risking short circuiting and corrosion. Where possible, re¬ ceptacle boxes should be raised up from the floor at least eight to ten cm. and the wiring enclosed in unbroken rigid conduit. It is also desirable to provide positive water drainage with floor drains spaced about six meters 22 [PDF page 25] FIPS PUB 31 apart. This is particularly important in new construction where the floor slab under the raised floor has been depressed to bring the raised floor flush with the surrounding floor. This eliminates the need for ramps but, with¬ out positive drainage in the depressed slab area, it is obvious that substantial amounts of water could collect under the raised floor. Not only would cables be submerged but each inch of water will add about five pounds per square foot to the live load, leading in extreme cases to structural damage or collapse. An increasing number of ADP facilities are now stockpiling plastic sheeting to protect ADP hardware in an emergency. Several cases have been reported where the prompt use of such sheeting has protected hardware against leakage from broken plumbing or fire fighting on upper floors. Because of the modest cost and assured effectiveness of this countermeas¬ ure, it can be recommended highly. 2.3. Earthquake Earthquakes represent a threat to ADP op¬ erations for two reasons. First, an earthquake may cause structural damage or collapse of the ADP facility building, interruption of electric or communications circuits, loss of utilities and other direct effects. Second are the more wide¬ spread effects on the community: disruption of transportation, food supplies and other vital services. As a result, many of the ADP staff may be unable to report for work and support¬ ing services may not be available. Assessing the probability of an earthquake is not easy because of the relatively short re¬ corded history of earthquakes in the United States. Figure 2 shows the number and in¬ tensity of known earthquakes and figure 3 is a seismic risk map based on these data. Note that the latter map merely indicates the probable severity, not probability of occurrence. On¬ going Federally-sponsored research is expected to lead to the ability to forecast long term probability and possibly even actual occur¬ rence. However, until such techniques become available it seems prudent for ADP facilities located in Zone 3 regions to assume that an earthquake which could disrupt operations for at least a week will occur at 50 to 100 year intervals. Furthermore, ADP facilities within about five to ten miles of major faults should probably assume total destruction of the fa¬ cility with about the same probability of oc¬ currence. There are two types of potential counter¬ measures. The first is to select a building with high resistance to earthquake damage and so located as to be protected against damage from neighboring buildings. Locations which should be avoided include hillsides, land fill areas, waterfront areas, fuel storage areas, tall struc¬ tures (such as buildings, radio towers or trans¬ mission lines) which might fall on the ADP facility or underground fuel transmission lines. One should bear in mind that the majority of the damage from the San Francisco earth¬ quake was caused by the subsequent conflagra¬ tion which raged uncontrolled from the lack of fire fighting water. For this reason considera¬ tion should be given to using sway bracing, flexible joints, etc. to make the sprinkler sys¬ tem earthquake resistant and to provide a re¬ liable on-site water supply. Beyond preventive measures such as these, the ADP security planner may wish to safe¬ guard the agency mission by including off-site operation in the ADP facility contingency plan. In this case he must be careful to select loca¬ tions which are sufficiently separated so as not to be affected by the same earthquake. Con¬ sideration should also be given to the location and construction of the facility used to store back-up files, documentation and the like in order to assure that these materials will be undamaged and accessible following an earth¬ quake. Valuable guidance in risk analysis and remedial measures will be found in “Building Practices for Disaster Mitigation” [59]. 2.4. Windstorms Windstorms, hurricanes and tornadoes all represent potential threats to an ADP facility. Hurricanes are characterized by high winds and heavy rain resulting in structural damage, flooding and in many cases loss of electric power. Of 148 major electric power interrup¬ tions in the United States reported during the period 1954 to 1966, 17 were attributed to hur¬ ricanes—an average of 1.3 per year. In 1970, Hurricane Celia was reported to have affected some 50 data processing facilities (some quite seriously) in the Corpus Christi area. Power was off for as much as 36 hours. A study of hurricane frequencies based on occurrences during the period 1886-1970, re¬ ported in “Atlantic Hurricane Frequencies Along the U.S. Coastline” [48], will be helpful to the ADP security planner in evaluating the exposure of his facility. Results of the study for high probability areas are summarized be¬ low : Annual Probability (Percent) Locations 16 Fort Lauderdale, Florida 15 Palm Beach, Florida 14 Brazoria County, Texas 13 Lafourche Parish, Louisiana 13 Mobile, Alabama-Pensacola, Fla. 13 Key West, Florida 12 Chambers County, Texas 11 Carteret County, North Carolina 9 Matagorda County, Texas 9 Franklin Parish, Louisiana 9 St. Bernard Parish, Louisiana 23 [PDF page 26] FIPS PUB 31 Figure 2. Damaging earthquakes of the United States through 1966. [PDF page 27] FIPS PUB 31 . "O *o .9? c5 a N c 03 t £Z — *o CD Qj ro ™ Q3 CO* u 2 S a | c E CO O (J o o > - S a)_ E 01 -a O = =3 g D 03 < a> rn 03 c CO c LU 03 _ "O , *> s s * co ■- p § Z E o ^ y 1 I O o » » X ^ O 03 o a o -o ~ ■Sep ■§ a o. c i— S ^ — OJ £ 3 S c P o • 2 O H3 m 7 ^ — 03 O 03 zz: 03 03 CD . O S E = *o £ E •O £ ra = E C O to ro a> n Q. 3 O -S’ C g a E ^ C E Jo .E -o -c c A § .co £ CO W OS P a fe ^ 03 X _ 03 03 25 [PDF page 28] FIPS PUB 31 Other localities on the Gulf and Florida coasts have probabilities in the range of 4% to 8 °/c. The probabilities for Atlantic coast areas not listed above range from 7% to zero. If the ADP facility is in or near the high prob¬ ability localities, the ADP security planner should give careful consideration to the threat from hurricanes. Apart from measures to protect against flooding and electric power failure, described elsewhere in these guidelines, one should con¬ sider the resistance of the ADP facility build¬ ing to wind damage, particularly windows broken by wind-driven debris or damage from falling trees, utility poles and the like. A “walk- around” inspection of the building should be adequate to identify potential trouble spots. Since ample warning is usually available, thought should be given to stockpiling plywood or similar materials for temporary protection of exposed windows and doors. The occurrence of tornadoes by state during the period 1953 to 1969 is depicted in figure 4. There was an average of 642 tornadoes per year. The mean number per 10,000 square miles per year is tabulated below for the high incidence states: Tornadoes/ State 10,000 Sq. Mi./Year Oklahoma 8.5 Kansas 6.0 Indiana 6.0 Massachusetts 5.4 Florida 4.9 Iowa 4.5 Nebraska 4.3 Missouri 4.3 For all other states the incidence is less than four. There is some evidence to suggest that tornadoes tend to reoccur in some relatively limited areas. Therefore one should not base an estimate of occurrence probability on the gross figures given above. Rather, if the ADP fa¬ cility is located east of the Rocky Mountains, the ADP security planner should consult with local authorities of the nearest National Weather Service office for information about the past record for the location of the ADP facility. tornado incidence by state and Figure 4. Tornado incidence by State (1953-1969). 26 [PDF page 29] FIPS PUB 31 During a recent tornado in Georgia a water main above the computer room of a data proc¬ essing facility ruptured and caused extensive flooding and the building evidently was badly damaged. Rapid reconstruction of the computer room in a nearby company building and hard work by the ADP staff and vendor engineers were major factors in rapid recovery. An effec¬ tive data base management system and central¬ ized administration of it, were also important factors. Some work was performed at off-site facilities and a week later work was back to the normal schedule at the temporary location. Even if there is no damage to the building itself, an ADP facility may lose electric power because of a nearby tornado. During 1954 to 1967 there were ten major electric power in¬ terruptions reported to be caused by tornadoes and seven more to be caused by high wind. To summarize, historic data should give a good indication of the probability of occurrence of hurricanes, tornadoes and high winds. Where the probability warrants the effort, the ADP security planner should give attention to measures to protect against building damage, flooding and electric power failure and should see that the contingency plan has the capability to meet such situations satisfactorily. 3. Supporting Utilities 3.0. Introduction Every ADP facility is dependent on support¬ ing utilities: electric power, air conditioning and often others such as communications cir¬ cuits, water supplies and elevators for its operation. The ADP security planner should consider the probability of occurrence and the effect of breakdowns, sabotage, vandalism and such accidents as fire, flooding and the like on these utilities. He can then relate the effects to the needs of the ADP facility as established by the risk analysis. This chapter discusses the factors affecting such events and measures to guard against them. 3.1. Electric Power Electric power as it affects ADP operations has two significant characteristics: quality and reliability. Quality is used here to refer to the absence of variations from the normal wave¬ form which are too small to be recorded by the local electric utility company but, depending on the ADP hardware, are large enough to affect operation of ADP hardware. Typically the ADP hardware rectifies the alternating elec¬ tricity, filters and voltage-regulates the result¬ ing direct current and applies it to the ADP circuitry. The filtering and regulation cannot be expected to eliminate voltage variations be¬ yond a reasonable range. If line voltage is 90% or less of nominal for more than four milli¬ seconds, or 120% or more of nominal for more than 16 milliseconds, one can expect exces¬ sive fluctuations in the DC voltage applied to the hardware circuitry. The effect on the circuitry is difficult to predict since it will depend on the amount and duration of the fluctuation and the state of the hardware. One may expect to find logic errors, erroneous data transfers or, in extreme cases, damage to hard¬ ware. Such things are usually obvious immed¬ iately, while other effects can go unnoticed until much later, if ever. These power line fluctuations, usually re¬ ferred to as transients, can be caused by light¬ ning strikes. Their probability of occurrence is dependent on the number of thunderstorms, the spacing between substations and the use of underground, as opposed to overhead, distribu¬ tion lines. Figure 5 shows the incidence of thunderstorm days in the United States. Ex¬ perience has shown that there will be approxi¬ mately one lightning induced transient at an ADP facility for every three thunderstorm days, with a somewhat higher rate in rural areas and about one third as many in urban areas where distribution lines are under¬ ground. Utility company transients are more difficult to predict but it is not unusual to find a tran¬ sient every morning at about 7 :30 a.m. when energy demand begins to build up and power factor correcting equipment is switched off¬ line. As a rule, such transients will not affect ADP operations, but cases have been reported where major problems were experienced every morning. Internally generated transients will depend on the configuration of power distribution in¬ side the building and the percentage of total load represented by the largest single switching load. The effects of internal transients can be minimized by isolating the ADP hardware from other building loads. Ideally the computer area power distribution panels should be con¬ nected directly to the primary feeders and should not share step-down transformers with other loads, particularly high horsepower mo- 27 [PDF page 30] FIPS PUB 31 THUNDERSTORM DAYS i i i i mar* 0 10 20 30 40 50 60 70 80 90 100 Lightning is the at¬ tendant of thunder¬ storms. The map at left shows the inci¬ dence of thunder¬ storm days—days on which thunderstorms are observed—for the United States. Alaska and Hawaii are less than 10. Figure 5. Thunderstorm days. tors. A typical power distribution system is shown in figure 12. This discussion has outlined the causes and effects of power line transients, but it is diffi¬ cult to develop good estimates for frequency of occurrence from abstract considerations. Fortunately, equipment is available which en¬ ables one to measure the actual occurrence of transients. Typically the device will include a strip chart recorder and electronic circuitry which will cause even brief or minor transients to be permanently recorded. By comparing the times when transients occurred with the con¬ sole log records of abnormal operation one can usually determine the number of disruptive transients in a given time period and often the cause of the transient. Such measurements should be made for at least a month and some ADP facilities do so continuously. However, there are two pitfalls. First galvanometer re¬ corders will not respond to brief transients and so display only the line voltage trend. For this reason they will not be helpful in dealing with transients. Second it is important to see that a qualified electrical engineer supervises meas¬ urements closely. If the measurements are to be useful, they must be carefully made, intelli¬ gently interpreted and correlated with other inputs. Discussions with representatives of the local electric utility will also be helpful in understanding the causes of observed tran¬ sients. The second basic quality of electric power— reliability—has to do with the number and duration of occasions when the line voltage de¬ parts from nominal for periods too long to be considered transients. One may observe sus¬ tained undervoltage (brownout) or actual fail¬ ure (blackout). Brownouts are a result of load near to or equalling generating capacity. In extreme cases the public utility will deliber¬ ately reduce line voltage by a maximum of 8% to stretch the generating capacity to meet de¬ mands. As a last resort they may actually disconnect a portion of the load, a procedure referred to as “load shedding,” but which, for affected customers, is a blackout. In addition, blackouts may result from windstorms, floods 28 [PDF page 31] FIPS PUB 31 and similar causes noted in Chapter 2, from failures of electric system equipment or, in rare cases, from human error. The famous Northeast blackout of 1965 re¬ vealed basic defects in the systems and pro¬ cedures for power pool management. Hope¬ fully, the measures since taken to increase the reliability of the national electric system make a repetition unlikely. Nonetheless, certain prob¬ lems remain, e.g., the inherent reliability of generating equipment, particularly very large units; and new probems are arising, e.g., en¬ vironmental protection measures, which make new construction to meet growing demand a lengthy process. The probability of occurrence of a blackout will depend on both random fail¬ ures at a more or less constant rate and the need for load shedding which depends on the amount of reserve generating capacity. Each factor must be evaluated separately. During the first half of 1967, fifty-two sig¬ nificant random power failures in the United States were reported by the Federal Power Commission (FPC) [10]. It seems reasonable to assume that this is a representative sample and that similar failures will occur at the same rate in the future. Less widespread or less significant events are not centrally reported —events such as transformer breakdowns, lo¬ cal accidents severing electric lines and other mishaps. There is no way to predict the fre¬ quency or imminence of these random or near¬ random events. The same FPC report suggests that the du¬ ration of randomly caused blackouts is about as follows: Duration Percent of Total Cumulative Total 9-15 minutes 6% 6% 15 - 30 minutes 36% 42% 30 - 60 minutes 18% 60% 1-2 hours 14% 74% 2 - 4 hours 10% 84% 4 - 8 hours 8% 92% 8-16 hours 6% 98% 16 or more hours 2% 100% The probability of loss of service due to blackouts or load-shedding by the local utility can be foreseen to some extent by becoming familiar with its generating capacity, its re¬ serves and, possibly, its current reliability and maintenance situation. If the reserve capacity is 20/7 of peak load, the probability of load related blackout is very small. As reserve ca¬ pacity approaches the capacity of the largest single generating unit, the probability of a blackout rises rapidly and an even lower re¬ serve capacity represents a precarious situa¬ tion. Current information in this and related areas can be obtained from FPC reports and the National Electric Reliability Council [22]. By considering all these factors, one can esti¬ mate the effect of power transients and failures with some confidence. By referring back to the risk analysis, he can then estimate the cost of these transients and blackouts to the ADP fa¬ cility. This cost estimate is then used to cost- justify protective measures. Of course, one should be careful to take into consideration projected growth in particularly sensitive ap¬ plications such as real-time or teleprocessing in projecting future loss potential. With a reasonable estimate of potential losses, the ADP security planner is in a posi¬ tion to evaluate candidate countermeasures on a cost-performance basis. There are a number of possible measures which address one or more quality problems at a range of costs. In the discussion which follows, general price ranges are included and will be stated in terms of kilovoltamperes (KVA) of load. While these prices will be helpful for preliminary analysis, they should be used with caution and final deci¬ sions should be based on accurate estimates. As a part of the analysis of protective meas¬ ures, the ADP security planner should obtain an accurate tabulation of these types of loads: the ADP hardware including data transmission devices, data conversion equipment, air condi¬ tioning equipment, normal and minimal light¬ ing and other equipment essential to emergency operation such as boilers, power doors, etc. He should make a “one-line” diagram of the electric power distribution arrangement for the building, particularly for the loads given above, down to the individual breaker panel level. These data are necessary to evaluate possible remedial measures to be described. If the major loss is expected to come from internally generated transients, a rearrange¬ ment of the power distribution may effectively solve the problem. No useful cost guidance can be given since it will depend on the particulars of the specific situation. In some cases it may be economically feasible to connect the building to more than one utility feeder via transfer switch. Thus if one feeder fails, the building load (or by splitting the main bus bar only critical loads) may be trans¬ ferred to the alternate feeder. This technique is of greater value if the two feeders connect to different substations. Since dual feeders only protect against localized blackouts, they are of limited value but one may in some situations find the cost justifiable. A voltage regulating transformer (VRT) will provide significant protection against mi¬ nor long-duration transients (4 milliseconds or more) and brownouts a^ a cost of about $100 to $200 per KVA of load. However, VRT’s will not protect against brief, high-intensity tran¬ sients or actual power failures. 29 [PDF page 32] FIPS PUB 31 At a cost of $200 to $300 per KVA, one can install a motor-alternator (motor-generator) set which includes an energy storage flywheel, as shown in figure 6. Such a configuration will protect very effectively against transients and power failures up to about 15 seconds in dura¬ tion. While reliability is quite high, one must allow for regular maintenance, particularly of bearings. It will be necessary to provide a spe¬ cial room for the equipment because the acous¬ tic noise level is quite high and the floor load¬ ing may be above normal. A number of vendors now offer what are referred to as uninterruptable power supplies (UPS). The typical UPS consists of a solid state rectifier which keeps a battery charged and drives a solid state inverter. The inverter synthesizes alternating current for the com¬ puter. A simplified block diagram is given in figure 7. In effect, the UPS simulates the motor-fly¬ wheel-generator set with the battery acting as a huge flywheel. Depending on the ampere- hour capacity of the battery, the UPS can support its load for as long as 45 minutes without input electricity. At the same time, it will filter out transients and compensate for brownouts. The cost for a UPS is in the range of $700 to $900 per KVA plus installation and site preparation costs, such as added air condi¬ tioning and floor reinforcement. To provide extra capacity, to clear load faults and to protect against a failure of the UPS, one can insert a static transfer switch between the UPS and the computer loads as shown in figure 8. The control circuitry for the static switch can sense an over-current condition and switch the load to the prime power source without causing a noticeable transient. When the total load exceeds 100 KVA or so, it may be economically feasible to use multiple, independent UPS units as shown in figure 9. Since each unit has its own disconnect switch, it can be switched off line should it fail for any reason. A.C. Input Figure 6. Typical motor-alternator set with flywheel energy storage. 30 [PDF page 33] FIPS PUB 31 A. C. Solid State D.C. Solid Stote A. C. Output Input Rectifier Inverter To Computer Loads Battery For Energy Storage Figure 7. Simplified block diagram of an uninterrupt able power supply. Solid State Transfer Switch Figure 8. UPS with transfer switch. 31 [PDF page 34] FIPS PUB 31 Figure 9. Multiple independent UPS units. Finally, if the risks analysis has shown a major loss from power outages beyond 30 to 45 minutes, one can install on-site generation, as shown in figure 10 at a cost of about $100 per KVA plus installation and site preparation. The prime mover may be a diesel motor or a turbine. When the external power fails, the control unit starts the prime mover automat¬ ically which in turn brings the generator up to speed. At this point, the UPS switches over to the generator. Barring hardware failures, the system will support the connected load as long as there is fuel for the prime mover. Note that the generator must be large enough to support other essential loads such as air condi¬ tioning, minimum lighting, etc., as well as the UPS load. There are many variations on the configu¬ rations shown here. If it appears that one or more of these measures can be cost justified, one should seek expert help in determining optimum performance specifications and the best overall solution to the problems of inte¬ gration into the building power distribution before deciding on a particular configuration. Furthermore, one must remember that in addi¬ tion to the rough cost guidelines given above, one must allow for any special installation costs, the cost of the floor space required for the equipment, the cost of any needed altera¬ tions to the air conditioning for the space, the cost for equipment maintenance and the cost of additional electric energy which will be dissipated by the equipment. Because of these complex cost factors, the analysis is a lengthy process. It is hoped that the discussion here will provide enough information to perfhit the ADP security planner to determine if a de¬ tailed analysis is warranted. A helpful discus¬ sion of UPS systems will be found in “Con¬ sultants Guide to Uninterrupable Power Sup¬ ply Systems” [57]. In the event of a fire, flooding or other emergency, it is important to be able to shut off electric power quickly, easily and selective¬ ly. First, one can use the power-off switch on the individual unit. However, one should re¬ member that the power cable and circuitry up to and including the built-in power-off switch are still energized. These can be de-energized by manually tripping the branch circuit break¬ er at the distribution panel. To do this easily and effectively, several conditions must be met. Distribution panels should be located in the computer room and access to them must be un¬ obstructed. It is not uncommon to find distri¬ bution panels hidden by other equipment or otherwise difficult to reach. Individual circuit breakers must be clearly marked so that one can quickly and accurately determine which circuit breaker is associated with each hard¬ ware unit. Finaly, one can disconnect all power 32 [PDF page 35] FIPS PUB 31 from computer room loads except for room lighting. While this can be accomplished by throwing the necessary disconnect switches, they may be located some distance from the computer room. To avoid this problem, RP-1 [9] requires that a master control switch be located near the console and just inside each principal entrance to the computer room which, when depressed, will disconnect power to all electronic equipment. NFPA Standard No. 75 [34] requires that power to ventilating equipment be disconnected as well, but it is suggested that this not be done without first considering the factors given in section 3.2. While these master control switches per¬ form a vital emergency function, it is obvious that their inadvertent operation will be ex¬ tremely disruptive. For this reason it is im¬ portant to see that they are clearly marked as to function and physically designed to re¬ quire deliberate effort to operate them. Figure 11. shows one solution to this problem. The master control switch shown in the figure is inside a plastic box located about six feet (2.0 m) above floor level. Accidental or careless op¬ eration appears to be highly unlikely. On- site Figure 10. UPS with on-site generation. Figure 11. Plastic box to protect master control switch against inadvertent operation. Photograph courtesy of Shell Oil Company 33 [PDF page 36] FIPS PUB 31 A one-line diagram of a typical building power distribution system is shown in figure 12 to clarify the preceding discussion. Begin¬ ning at the top, we see that power flows through a series of step down transformers, disconnect switches and overcurrent protective devices (fuses) until it reaches the individual distribution panels. Each panel has a number of circuit-breaker protected branch circuits to which individual hardware units are con¬ nected. This basic configuration can be modi¬ fied in a number of ways to enhance quality or reliability. First one could take pains to iso¬ late ADP circuits from equipment which gen¬ erate transients, e.g., high horsepower motors. The greater the distance from the ADP facility to the substation, the greater the probability of a feeder failure, all other things being equal. If feeder failure appears to be a signicant threat, one can usually arrange for a second feeder (ideally from a different substation) to be run to the ADP facility. A transfer switch which can be either manual or automatic is used to switch the step down transformer from the primary feeder to the back-up feeder in the event of a failure. Alternatively, one might isolate critical building loads, e.g., ADP bus bar, ADP air conditioning, emergency lighting, security hardware, and supply them through a completely separate power distribution sys¬ tem. In this case only the critical load need be switched to the back-up feeder. This ar¬ rangement insures isolation, and the cost of the back-up feeder is reduced since it does not have to carry the entire building load. This may have a major impact on the cost justifi¬ cation. With the help of the building manager or engineering staff, the ADP security planner should check these points about the power dis¬ tribution system: (a) Electric wiring conforms to the require¬ ments of the National Electric Code [55], NFPA No. 75 [34] and RP-1 [9], (b) Procedures are established in coordina¬ tion with the building manager to insure that electrical maintenance work is coordinated with ADP operations to avoid inadvertent shut-off of computer room, air conditioning or communications power. It may be desirable to label sensitive disconnect switches “up stream” of the computer room, but not in such a way as to flag them for a saboteur. (c) All electric power distribution equip¬ ment is adequately protected physically against accidental damage or sabotage. Pro¬ tection may include such things as control over access to electrical equipment rooms and clos¬ ets, barriers to protect utility poles and ex¬ terior transformer pads against damage by vehicles and avoidance of proximity to fire hazards. In summary, the appropriate steps should be taken to assure that the quality and relia¬ bility of electric power will satisfy the needs of the ADP facility. Depending on the risk analysis and cost factors these measures may include changes to the power distribution sys¬ tem configuration, dual feeders, devices to filter out transients, uninterruptable power supplies, devices to compensate for brownouts, on-site generators and physical protection against tampering, sabotage or accidents. In addition, the wiring should conform to applicable codes and be properly integrated with the fire safety program. 3.2 Air Conditioning Properly conditioned computer room air is important for three reasons. First the elec¬ tronic circuitry requires fairly close tempera¬ ture limits to minimize erratic operation. High temperatures (above about 30°C) may cause permanent damage to ADP hardware. Second, humidity control is required to assure proper operation of tabulating card devices and tape drives. Excessive humidity may cause cards to swell and feed erratically. Very low humidity often leads to static electricity buildup which can affect tape handlers, line printers and sometimes the ADP hardware itself. Finally, it is important that the room air be free of contamination which may be corrosive, con¬ ductive or large enough to cause disk drive head-crashes.* To the extent that controls over temperature, humidity or contaminants fail, ADP operations may be hampered or hard¬ ware damaged. In extreme cases it may be necessary to suspend operations until the situa¬ tion can be corrected. Furthermore, if the computer room is a part of a building-wide air conditioning system, smoke from a fire else¬ where in the building may be introduced into the computer room. In order to properly assess the exposure to these potential hazards, the ADP security planner should review the air conditioning sys¬ tem for the ADP facility with the building manager. Figure 13 shows a typical air condi¬ tioning system in diagrammatic form. The heart of the system is the air handling unit (AHU) through which computer room air is circulated by a fan. The function of the AHU is to provide temperature and humidity control and air filtering. To refresh the room air, out¬ side air is drawn in through a louver in an exterior wall and mixed with return air. In ad¬ dition, there may be an exhaust fan as well. * One type of humidifier operates by atomizing water and injecting it into the air stream. This type should not be used in hard water areas because minerals in the water will be deposited on the ADP hardware. 34 [PDF page 37] FIPS PUB 31 Connection to Local Utility Feeder Figure 12. Simplified one-line diagram of power distribution. 35 [PDF page 38] mm\ w\ FIPS PUB 31 Heat to Air Wl Heat Exch. % Heat Pump Y77\ '// ^ Exterior Wolls Circul. Pump ( FiIte r Air Handling Unit ✓ Ducts N Outside Air- 7 Control Dampers * 0 £_L / 000M 1 Fan Reheat Cooling Humidification r Fan Return Supply J Exhaust Air Air Air Heat to Room Air .... Louve A D P Hard wa re Computer Room Figure 13. Cross-sectional schematic diagram of a typical air conditioning system. 36 V////JV////P////////A [PDF page 39] FIPS PUB 31 Air flows through ducts, usually made of sheet metal, and proportioning is controlled by mo¬ torized dampers. To perform its function, the AHU needs a supply of water or steam for hu¬ midification during periods of low humidity, and some way to exhaust the heat removed from room air. This latter is done by connect¬ ing the AHU with some kind of heat pump (a chiller, direct expansion unit, etc.) by means of a refrigerant (e.g., chilled water) circulated by a pump. Likewise, the heat pump must have some means to dissipate the heat, usually a cooling tower or condenser. The actual arrangement of system elements will depend on its size and local conditions. For example, a typical residential window air conditioning unit will combine all the functions except humidification into a single unit. Quite often computer rooms make use of so-called packaged air conditioning units which perform all functions except air intake and exhaust and heat exchange. In large buildings it is quite common to use one or a few heat pumps to sup¬ port building comfort air conditioning as well as computer room AHU’s. From this discus¬ sion one can see that there are many different devices which can fail with different conse¬ quences to ADP operations. The major failure modes, their effect and possible countermea¬ sures are tabulated below in general terms. Failure Effect Countermeasures Outside air damper or fan. No outside air, but usually not critical. Multiple outside air sources. AHTT fan. No air circulation. Temperature rises. Multiple AHU’s. AHU humidity control. Loss of humidity control. Critical if outside air hu¬ midity is very high or low. Multiple AHU’s. AHU temperature control. Temperature rises. Multiple AHU’s. Circulating pumps, heat pump or heat exchanger. Temperature rises. Multiple units in¬ terconnected so affected unit can be taken off line. Use outside air, and even floor fans, temporarily. To minimize the effects of failures, one can use multiple units, interconnected to permit affected units to be taken off line or to permit outside air to be used in an emergency. As an example, consider the situation where the com¬ puter room requires 50 tons of cooling, the bal¬ ance of the building requires 100 tons for comfort air conditioning and a chilled water system is to be used. Two different system con¬ figurations are tabulated below: Simple One 150 ton chiller One chilled water circulating pump One 50 ton computer room AHU Redundant Three 50 ton chillers Three chilled water circulating pumps Three 20 ton computer room AHU’s While the simple system will meet the need, the failure of any single piece of equipment will probably require ADP operations to be halted within a few minutes to a half hour. The redundant system will be somewhat more expensive but failure of a given unit can be accommodated. If one or two chillers or circu¬ lating pumps fail, the computer room can still be supported by reducing or cutting off the comfort air conditioning to the balance of the building. If a computer room AHU fails, oper¬ ations can probably be continued by reducing the heat load. This can be done by reducing lighting and turning off the least important ADP hardware. Both as an emergency procedure and as nor¬ mal energy conservation, outside air can be used for cooling if the temperature and hu¬ midity are low enough. How high the tempera¬ ture of the outside air may be and still be effective for cooling depends on three things: the maximum allowable room-ambient or equipment intake temperature (either or both may be specified), the amount of heating that takes place in the air-handlers and ducts and the degree to which outside air (as opposed to recirculated warm air) may be used. Most of the existing air conditioning installations do not allow for an intake of only outside air, although in some cases it may be feasible and cost effective to modify the ducts and venting to permit this. Assuming a 100% intake of outside air and exhaust of room air, there can be a tempera¬ ture rise of up to 15 °F (8 °C) between the temperature at the intake to the air handling units and the warmest spot in the computer room. If, therefore, the maximum allowable temperature in the computer room is to be 90 0 F (32 °C), then the highest temperature at which outside air may be used would be on the order of 75 °F (24°*C). However, this should be determined for each installation, based on its equipment specifications and air condition¬ ing configuration. In extreme emergencies it may be possible to use floor fans to exhaust computer room air to other parts of the building. To evaluate the inherent system reliability, one should consider the factors already dis¬ cussed, past failures and the estimated time to repair. This latter will depend on the availa¬ bility of spare parts and qualified service per¬ sonnel. The building engineering staff will be able to help with this estimate and with con- 37 [PDF page 40] FIPS PUB 31 sideration of alternate means of increasing reliability. It is also desirable to keep one or more temperature-humidity recorders to moni¬ tor performance. Assuming normal operation, such records should be reviewed each week to discover erratic or inadequate performance, identify the cause and institute corrective ac¬ tion. One recorder should te kept in a fixed, central location to permit week-to-week com¬ parisons. Additional fixed units may be desir¬ able for computer rooms in excess of 1,000 square feet (100 m.2). Finally, if problems are encountered with even temperature distribu¬ tion, it may be helpful to have an additional recorder for spot checking. Since computer hardware is relatively sensi¬ tive to dirt and corrosion, the source and filter¬ ing of the outside air is important. When air intake louvers are located at ground level, there is a danger that excessive dust or dan¬ gerous fumes will be ingested. In one case a skunk near an air intake louver was disturbed by a maintenance worker who was cutting the grass. The resulting odor forced the total evac¬ uation of a three story building! Further, it is important to see that filters are adequate and that they are inspected regularly and cleaned or replaced as needed. Because the air conditioning system is used to move air within the building, it is important to be able to predict and control its operation during a fire. Referring to figure 13 one can see how the air conditioning system can be used to exhaust smoke from a computer room by closing the return air damper and fully opening the intake and exhaust dampers. Since prompt smoke removal will limit damage and permit fire fighting, such an arrangement is preferred to a complete shut down of air condi¬ tioning. However, if smoke will be forced into other parts of the building or ducts will be sub¬ jected to high temperatures, then shut-down is required and can be included as a part of the functions of the master control switch de¬ scribed in section 3.1. Figure 14 shows a typical building air con¬ ditioning system. Return air and fresh air are mixed at the top floor of the building, passed through an AHU and then distributed to each floor of the building via the main supply duct. It can be seen that with such a system, smoke from a fire on the first floor would be quickly distributed throughout the building unless fire dampers were provided. Furthermore, the duct work may provide an avenue for the spread of a fire. In a recent ADP facility fire, air condi¬ tioning ducts were routed along the basement ceiling and then up through holes in the floor slab to a first floor computer room. When a fire started in packing materials stored in the basement, these ducts quickly failed and heat and flames entered the computer room. Ex¬ tensive damage was done to hardware and supplies. For these reasons, air conditioning systems should conform to NFPA Standard No. 90A [28] as required by RP-1 [9], Figure 14 illustrates a number of these requirements which can be briefly summarized as follows: Where ducts pass through fire walls they are equipped with automatically operating fire doors. Fire dampers are required at fire rated walls which are intended to restrict the spread of the fire, at openings in vertical shafts and other similar points. Smoke and heat detectors properly located in the duct work and emergency shut down controls are required to protect the system against smoke or high temperature air. Figure 14. Typical air conditioning system in a fire resistive building. 38 [PDF page 41] FIPS PUB 31 NFPA Standard No. 90A [28] also requires that ducts, filters and other parts be noncom¬ bustible, that electrical wiring and equipment conform to the National Electric Code [55] and that in general the air conditioning system not defeat building features intended to limit the spread of a fire. The Standard also includes criteria for determining if the system can be used safely for smoke removal as has been suggested above for the computer room. The key factors are the ability of the system to handle high temperature gases and the effect on life safety objectives. In summary, it is im¬ portant for the ADP security planner to un¬ derstand the operation of the building air con¬ ditioning as it effects fire safety, and to identify the corrective actions needed to pro¬ vide protection for the ADP facility. In section 3.1 it was asserted that emergency electric gen¬ erating equipment should have enough capa¬ city for minimum lighting and air conditioning as well as for the ADP equipment. It follows that the efficiency of the air conditioning sys¬ tem then effects not only its own cost of opera¬ tion, but also the size and cost of emergency generators. The power required to operate ADP air conditioning is substantial, being on the order of 40% to 75% of the power re¬ quired by the ADP equipment, lighting and other loads. This says that for every kilowatt of load removed in an emergency, the power input requirement is reduced by roughly one- and-one-half kilowatts. Few ADP air conditioning systems were de¬ signed with energy costs and unavailability and the requirement for backup electric gene¬ rators as significant design constraints. This may be one of the reasons it is quite common to find computer rooms operating at 72-75 °F (22-24 °C) and 50% RH and consequent dew¬ points of 52-55 °F (11-13 °C), while the chilled water used for cooling may be supplied at 42 °F (6° C). Therefore, the chilling units are constantly extracting water from the air. Not only does this reduce the cooling efficiency and require considerably more energy, but even more energy will be required to add water back into the air to bring the relative humidity back to 50%, generally done by injecting steam (which in turn counteracts the cooling). In ex¬ isting installations, energy savings may be ac¬ complished by lowering the relative humidity, by lowering the computer room temperature (particularly when recirculated air is being chilled) or by raising the temperature of the chilled water (where the savings appear in re¬ frigeration-compression costs). In new facili¬ ties, the need for emergency electric generators and increased fuel costs can be factored into the original design to achieve an optimum solution. None of these suggestions should be under¬ taken without a thorough evaluation by heat¬ ing and air conditioning specialists through GSA or the building engineer for privately owned facilities. Equipment manufacturers should be consulted if one anticipates lowering the relative humidity significantly, such as to below 35% RH, because of the possibility of static electricity problems. 3.3. Communications Circuits Increasingly ADP systems are making use of communications circuits for rapid data entry and output. It is important to see that the re- liabiliy and integrity of the communications circuits satisfy the requirement of the ADP fa¬ cility. Figure 15 shows ,a representative tele¬ processing equipment configuration. A specific teleprocessing system may use any one or more of the elements shown in figure 15. As a rule there will be some identifiable hardware unit or units (referred to here as the message proces¬ sor) which acts as the interface between the computer and the circuits to the individual terminals. Circuits may be hardwired DC cir¬ cuits or may use modems as shown in the figure. A terminal may be “stand alone”, using either a leased line or the dial-up network for access. It may be one of several terminals (usually at several locations) which share a multi-drop leased line or one of several low speed terminals (usually at the same location) which share a high speed leased line via a con¬ centrator. Typically the configuration has been selected to minimize the total direct cost taking into account the cost per minute of dial-up calls, monthly charges for leased lines of WATS lines and lease or capital costs of terminals, modems, etc. However, the cost of delays re¬ sulting from communications failures raay be significant and provide justification for the di¬ rect cost of measures to increase reliability. If the risk analysis has indicated a significant loss potential from delayed processing, the ADP se¬ curity planner should attempt to estimate the rate and duration of failures and look for re¬ medial measures which can be cost justified. The following are some of the potential failure modes: One channel of the message processor, one local modem or one telephone circuit to the local central office fails. The result is one chan¬ nel out of service until the failed element is repaired and, if the channel was in operation at the time of the failure, one incompleted message transmission. If access is via the dial¬ up network, remote terminals can still access the ADP system, although there may be in¬ creased waiting time during busy periods. If access is via leased lines, only the remote ter¬ minal (s)* connected to the failed circuit will be affected. A message processor circuit failure cannot be overcome until the unit is repaired * Note that failure of a multi drop or concentrator circuit will affect more than one terminal. 39 [PDF page 42] FIPS PUB 31 Figure 15. Representative teleprocessing equipment configuration. 40 [PDF page 43] FIPS PUB 31 unless there are space circuits one of which can be quickly substituted for the failed circuit. In general it should be relatively easy to re¬ place a failed modem with a stand-by unit. Re¬ pair of a circuit to the local central office will probably be completed within a few hours in most cases but the risk analysis may indicate the need for one or more spare circuits. The entire message processor, all circuits to the local central office or the central office itself may fail. Any of these result in cutting- off all remote terminals and messages being- transmitted. A message processor failure is probably the most likely of the three and the repair time may be quite protracted. The ADP security planner should consult with the ven¬ dor, review the past history of the unit and attempt to estimate the probable failure rate, and mean time to repair. If the risk analysis supports the cost and it is technically feasible, one may elect to install multiple units which share the common traffic load so that the fail¬ ure of a single unit will not be catastrophic. By consulting with representatives of the local telephone company or Federal Telecommunica¬ tions System, the ADP security planner can de¬ termine the practicability of installing a sepa¬ rate set of circuits to another central office. While the probability of the simultaneous fail¬ ure of all circuits to the central office is quite low, it is not zero. In June, 1973 it was reported that thieves had cut the telephone cable leading into a central burglar alarm station. They then broke into and robbed several of the protected premises before the cable could be repaired. This points up the potential exposure to sabo¬ tage or vandalism. Cables are also exposed to construction excavation, ice storms, utility pole knock-down, manhole explosions, floods, dam¬ age from fires inside the building and earth¬ quakes. Failure of a circuit from the local central of¬ fice to a remote terminal or of the terminal itself or its modem. This is the least critical failure since it affects only one terminal and does not impinge on ADP operations. The time to recover from a circuit failure will usually be a few hours for a leased line. Terminal or modem repair time will depend on availability of vendor service support. Based on an estimate of the expected failure rate of the terminal and modem and the mean time to repair, the ADP security planner and the terminal users can project the associated loss potential and so de¬ termine if standby equipment can be cost justi¬ fied. Unless there are many terminals at the remote location or the application is particu¬ larly time-sensitive, standby equipment prob¬ ably will not be justifiable. This outline analysis of failure modes leads to several points which the ADP security planner should consider: Dial-up versus leased lines. As a rule one se¬ lects leased lines when the amount of traffic reaches the point where leased lines are less expensive than dial-up toll charges or condi¬ tioned lines are required because of the data transmission rate. Leased lines lack the inher¬ ent reliability and flexibility of the dial-up net¬ work. The time urgency of some user applica¬ tions may justify the cost of additional leased or dial-up lines for back-up. However, it will be necessary to provide the hardware (line switching), software and operating procedures to make full use of back-up lines. Finally, dial¬ up exposes the system to foreign terminals. At the same time he is investigating the re¬ liability and mean time to repair of communi¬ cations circuits, the ADP security planner should examine means to restore communica¬ tions at an alternate site in the event of a catastrophe. The ADP security planner should also consider alternate means to process user input and output, e.g. use of other remote ter¬ minals or on-site input-output devices. This in¬ formation is a vital input to the development of the back-up planning described in Chapter 8. The ADP security planner should examine the way in which the teleprocessing software handles failures. The key points are: • Recognition of a failure and generation of helpful diagnostic messages at the con¬ sole. • Proper handling of interrupted messages particularly as they may affect file updates • Software flexibility to accommodate a failed channel and the reassignment of users to alternate channels or terminals • Alternate software to accommodate back¬ up modes of entry and output. Finally one should examine the security of communications circuits. Terminal boards and other equipment should be located in locked rooms to which access can be controlled. Cables should be so routed as to protect them against physical damage, preferably by placing them in rigid conduit. Procedures should be established to coordinate telephone system changes and repairs. Care should be taken to show the loca¬ tion of underground cables accurately on ADP facility site drawings and to assure that subse¬ quent excavations are properly planned and su¬ pervised to avoid cutting cables by mistake. Communications circuits are also subject to more subtle tampering. A 1971 newspaper re¬ port* describes alleged sabotage of a system by * "System Sabotaged by Phone". Computerworld, p. 1. De¬ cember 15, 1971. 41 [PDF page 44] FIPS PUB 31 a group of strikers. According to the report, computer polling commands were tape recorded and then transmitted via the dial-up network to remote terminals. The result was to prevent subsequent polling of the terminals by the com¬ puter. Twenty-five terminals were affected for nearly a month. This episode suggests the pos¬ sibilities for what might be called software sabotage. The ADP security planner should re¬ view communications software and procedures and if there is significant exposure to tamper¬ ing, identify modifications which will reduce the exposure, insure rapid discovery and mini¬ mize potential damage. Wire tapping; message intercept, alteration and forwarding; access by an unauthorized user via the dial-up network and other aspects of controlled accessibility are not included in this handbook. 3.4. Other Supporting Utilities Electric power, air conditioning and commu¬ nications are clearly vital to ADP operations, but other utilities may also be required for normal operations. These are some possibili¬ ties which should be examined: Water supply. Because water is probably re¬ quired by the air conditioning system and the heating plant, the loss of water pressure may halt operations. A temporary loss of water for drinking and fire fighting purposes probably will not interfere with operations immediately. Water may also be required for processing of microfilm or other photographic media. Elevators, particularly in high rise buildings, may be important for the movement of people, data and supplies. It is unlikely that all eleva¬ tors will fail simultaneously except in the event of an electric power failure. However, if it is essential to keep one elevator operating, one must provide an on-site generator which may, of course, also be required for the ADP hard¬ ware. In some large facilities internal mail convey¬ ors or pneumatic tubes may be used to deliver source documents or output. It is likely that hand delivery can be substituted, if necessary, but the ADP security planner should verify this and also consider if urgent material in transit at the time of the failure might be trapped in the equipment. In a few cases, building heating or air con¬ ditioning may be supported by steam generated outside the building. If this is the case, the ADP security planner should investigate the reliability of the source and the effect of a failure to determine the possible need for alter¬ nate sources or for special provisions in the contingency plan. Building heating or air conditioning may de¬ pend on natural gas supplied by a public utility. The considerations are much the same as with an external steam source. If an uninterruptible supply is found to be important, the risk analy¬ sis may provide cost justification for an on-site back-up supply. To analyze these and related matters and to examine the cost and feasibility of counter¬ measures, the ADP security planner should seek qualified professional help from the build¬ ing manager and other technical specialists available to him. Because of the interrelationship of heating, air conditioning and electric power, a number of recent buildings have made use of what is referred to as a total energy system (TES). Stated simply, a TES integrates these elements into a single system to provide all three func¬ tions. Typically electric power is generated on site and exhaust heat is used for building heat¬ ing. It has been reported that the overall cost can be less than separate systems and one has the advantage of control over the source of energy. This means that reliability and quality can be tailored to specific user needs. For these reasons, it is worthwhile to considei a TES where planning a new facility, but the ADP security planner should apply the same stand¬ ards for quality and reliability as he would to conventional systems and be sure that ADP facility requirements will, in fact, be satisfied. 4. Computer System Reliability 4.0. introduction Without question, computer reliability is fundamental to ADP operations. However, computer reliability does not always receive adequate attention, often because responsibility for it is not clearly assigned. This chapter in¬ troduces three basic areas—reliability of ex¬ isting computer systems, maintenance man¬ agement, and procurement of new systems— and suggests ways to deal with them. 4.1 Computer System Reliability The typical computer is composed of many interconnected units which perform the func¬ tions necessary to complete assigned data pro¬ cessing tasks. In the simplest situation, the computer performs a single task and so would probably be configured to use the minimum number of hardware elements required by the task. Thus, the failure of any element would halt operations. In the more typical multi-task 42 [PDF page 45] FIPS PUB 31 environment not all tasks will use all the re¬ sources of the computer, and so a failure will not necessarily prevent completion of all tasks. Most computers use an operating system to control the job stream and to allocate memory and peripheral devices to individual jobs. De¬ pending on its features, the operating system will detect failures as indicated by hardware alarms, attempt to localize and define the fail¬ ure, notify the console operator and adjust its control of the job stream to maximize the num¬ ber of tasks which can continue to be executed. Of course, failure of the central processor control logic will usually halt operations. (Note, however, that a failure might go undetected and could disable hardware which controls access.) Likewise the failure of one-of-a-kind peripheral units will interrupt all tasks which use them. Thus failures may permit all tasks to be performed but at a lower throughput rate, may prevent the performance of some tasks or may completely halt operations. In order to understand the impact of hard¬ ware failures on the reliability of ADP opera¬ tions, the ADP security planner should conduct a system failure mode study by examining the impact of each significant hardware failure. He can do this by noting the computer system re¬ sources required by each of the applications identified by the risk analysis as time critical. If the system is at all complicated, he will prob¬ ably want to consult with staff members re¬ sponsible for the hardware and operating sys¬ tems and the vendor’s technical support per¬ sonnel. The typical ADP procurement will include standards of performance demonstration re¬ quired for acceptance of a system. Review of the acceptance test documentation will often be helpful to the ADP security planner in estimat¬ ing system reliability of an existing installa¬ tion and identifying units most likely to fail. The objective is to use the failure mode analysis, the loss potential of urgent tasks, and estimates of failure rates and repair times, for projecting future losses to the ADP facility from hardware failures. The projection will permit the ADP security planner to identify those hardware units where failures will be most critical to operations as the basis for the cost justification of remedial measures, as a guide for development of a contingency plan and as an aid in future procurement decisions. If the analysis shows a significant loss poten¬ tial from hardware failures, the ADP security planner can consider the following alternatives: • Incorporate one or more additional units of a given type beyond the minimum required to perform the stated task load to permit con¬ tinued operation in the event of the failure of a unit when the analysis shows it to be critical. • Alternatively, eliminate a critical periphe¬ ral unit and substitute an alternate technique or procedure. In other words it may be possible that the savings in operating cost resulting from use of a specialized input device might be outweighed by the exposure to losses caused by its failures. • Take steps to reduce failures and speed¬ up repairs as described in section 4.2. • Install two or more computers which as a group can handle the normal work load. If one computer fails, only the least critical tasks will be interrupted. • Install two (or, indeed, several) identically configured computers so that either system can perform all assigned tasks. While this approach (dual or multiple computers) will be difficult to cost justify in most cases, it may be the only acceptable solution for extremely critical or high risk missions. 4.2. Management of Hardware Maintenance Apart from optimizing the system configu¬ ration in terms of achieving established reli¬ ability goals, it is important to establish ade¬ quate policy and procedures for management of hardware maintenance. Effective mainte¬ nance management should include these activi¬ ties: • Determine the optimum schedule and scope of preventive maintenance; arrange for on¬ going supervision to reduce failures to an ac¬ ceptable level, if possible. As a rule, provisions for preventive maintenance will follow the ap¬ plicable Federal Supply Schedule but can be modified by mutual agreement between the vendor and the government. • Report and perform statistical analysis on hardware failures so as to detect significant failure trends and take remedial measures on a timely basis. This implies that ADP Opera¬ tions Branch must report all system failures in enough detail to permit the technical staff to determine the cause of the failure. One ADP facility uses the following procedure: When¬ ever the system goes down regardless of the apparent reason, a System Incident Report (SIR) is prepared by Operations. The SIR form calls for full information including the time of day, system status, tasks and jobs in the sys¬ tem, diagnostic messages, availability of core dumps and the like. The form also provides spaces for information about routing of the SIR and the final disposition of the incident. At the same time, the incident is added to a log of unresolved incidents by ADP Technical Services Branch. When the incident appears to 43 [PDF page 46] FIPS PUB 31 be caused by hardware, a vendor representative is notified immediately. When the cause is soft¬ ware or unknown, the SIR is passed to the Current Systems Branch for disposition. When the cause of the incident has been discovered, the appropriate agencies take corrective ac¬ tion as needed. The SIR is completed, copies with supporting documents are disseminated to appropriate functions, and the log entry is closed out. This or a similar procedure will insure that problems are discovered and dealt with effectively and that the needed informa¬ tion about system operation is retained. At regular intervals, the ADP staff member assigned responsibility for system reliability should analyze these reports to identify unfa¬ vorable trends. Careful maintenance of mean¬ ingful, detailed reports can be of great value. Without them an unfavorable hardware trend may go unobserved for an unnecessarily long period of time and identification of the cause may be further delayed while specific informa¬ tion is being acquired. Full use should be made of error reporting features available in the op¬ erating system. • Remedial maintenance should also receive continuing attention. The analysis of loss poten¬ tial associated with hardware failures may show that efforts to reduce the mean time to repair may be particularly cost effective. Pro¬ visions for remedial maintenance are specified in the applicable Federal Supply Schedule but, where the need can be supported by an analy¬ sis, the ADP facility may elect to arrange for on-site maintenance personnel or stock piling of critical spare parts. 4.3 Reliability Considerations for New Systems It is not unusual to find that inherent system reliability receives little detailed consideration in the design of a new system. Paragraph 101- 32. 402-7 of the Federal Property Management Regulations [15] defines data system specifica¬ tions in part as including “. . . a description of the data output and its intended uses, the data input, the data files and record content, the volumes of data, the processing frequen¬ cies, timing and such other facts as may be necessary to provide for a full description of the system.” What is suggested here is that “such other facts” should properly include a consideration of reliability. The typical Federal Supply Schedule (FSS) will call for a System Effectiveness Ratio (SER) (operating time divided by operating time plus failure down time) of 90%. It is likely that the system designer accepts this figure for throughput estimates with the realization that work load and run time estimates are of compa¬ rable accuracy. Experience suggests that an SER of 90% will be acceptable for the typical batch mode operation but on-line service re¬ quires an SER of at least 95%. If the ADP system is involved in life support in any way, a much higher figure is probably required. The ADP security planner should look closely at the characteristics of the planned and likely future work load to test the validity of the system reliabiliy assumptions. Notice that the SER is roughly equivalent to mean time between fail¬ ures (MTBF) divided by MTBF plus mean time to repair (MTTR). thus if 160 hours are scheduled for a week (20 eight-hour shifts) one could have eighteen hours of down-time and still achieve a 90% effectiveness ratio. If the nature of the projected work load would make this much down-time unacceptable, addi¬ tional consideration of reliability is in order. The measures already described for existing systems (dual systems, redundancy within a system and accelerated repair) apply to new systems as well, but it may be easier to apply the first two during system design than after installation. In cases where the ADP hardware has had significant use elsewhere, it may be possible to get more realistic figures for MTBF and MITR from the vendor. In such cases, the re¬ liability analysis will be benefited even if it is not appropriate to include reliability figures as contractual requirements. This will be particu¬ larly true in the cases where the reliability of a system appears to be marginal based on a 90% SER but corrective efforts cannot be cost justified easily. If credible vendor-supplied esti¬ mates indicate that a higher ratio will, in fact, be achieved, one might conclude that system reliability will be acceptable. Finally one should note that SER does not indicate the duration of hardware failure inter¬ ruptions. Continuing the example above, one might have one 18-hour interruption per week, six 3-hour interruptions or any other combina¬ tion limited only by the response time of serv¬ ice personnel. For this reason, the ADP secur¬ ity planner should attempt to determine the likely distribution of interruption durations and examine the implications on performance of urgent tasks since six 3-hour interruptions might not cause any significant loss, but a single 18-hour interruption could be quite serious. ( 44 [PDF page 47] FIPS PUB 31 5. Physical Protection of ADP Facilities 5.0. Introduction This chapter addresses the requirements for physical protection of the ADP facility which can be thought of as the process of permitting access to the facility by authorized persons while denying access to others. It is helpful to think about the problem in three dimensions: the roles of people, e.g., computer room opera¬ tor, ADP programmer, vendor representative; the criticality of specific areas, e.g., the sur¬ rounding grounds, public areas inside the build¬ ing, mechanical equipment rooms, the tape li¬ brary; and the time of day, e.g., normal busi¬ ness hours, computer room second and third shifts, periods when the ADP facility is unoc¬ cupied. The objective of the physical protec¬ tion plan is to establish go/no-go criteria for all combinations of these three dimensions and then provide measures to implement them. In other words for each class of individual, the times for which access is permitted is stated for each specified area. To develop these go/no- go criteria, the ADP security planner should conduct a systematic and comprehensive an¬ alysis of the threats to which the ADP facility is exposed, the physical characteristics of the building which houses the ADP facility and the organization and mission of the ADP facility. Since the physical protection and controls over access by people will cost money to implement and operate and may represent some impedi¬ ment to work flow, it is important to try to achieve the optimum level of protection . . . neither inadequate to achieve stated security goals nor needlessly expensive or cumbersome. Likewise it is important to have balanced pro¬ tection against all determined risks. A senior bank officer recently observed that there was a tendency to build . . . steel doors in paper walls,”* a very graphic description of unevenly applied security measures. For just such rea¬ sons the effort to determine protection needs on a realistic basis is well worth the effort. 5.1. Determining Protection Requirements The first step in the determination is to evaluate the potential threat to the ADP fa¬ cility from outsiders. Since one is dealing with human motivation there is no easy way to be qualitative. However, one should attempt to make a reasonable determination for each of the classifications which follow. Specifically, consider how both the ADP facility and build¬ * ''IBM, Security Test Sites Vie on Software Strength”. Computerworld, p. 1, June 13, 1973. ing tenants will appear to attackers. While determining the likelihood of attack, one should also estimate the likely level of effort the wrongdoer might be willing to exert to achieve his goal. Common criminals. The concern here is with theft of government property. Would a burglar be likely to think there is valuable property in the building? This might include office ma¬ chines, firearms, drugs, cash, personal posses¬ sions or any other items subject to easy resale or useful for other criminal activities. Activists. Is the agency active (or thought to be active) in fields which are controversial? Might the building be thought of as a desirable symbolic target at which to direct attention getting demonstrations. An activist group forced entry at a midwestern research labora¬ tory’s ADP facility with the intention of de¬ stroying magnetic tape data files for research projects of which the group disapproved. No employees were present at the time and the activists did not damage any of the hardware. A number of tapes were said to have been erased and punched cards and the like were thrown on the floor. The group was not dis¬ covered during the break-in but revealed them¬ selves at a press conference a few days later. The research laboratory is said to have in¬ creased its patrol force coverage and given consideration to intrusion detectors subsequent to the break-in. While damage was estimated to be no more than $100,000 this episode points up the importance to safeguarding an ADP facility against intrusion. Espionage agents. Does the ADP facility hold or process data which could be of value to an outsider prior to its public release such as eco¬ nomic activity, future allocations of Federal funds or sensitive personal information? Vandals. Is the ADP facility located in an area where vandalism is prevalent? The second step in the analysis is to define and tabulate areas within the facility for con¬ trol purposes. The tabulation should include a statement of the location, function, access requirements (what people at what times), and criticality (contents or activities which may be targets for wrongdoers) for each area. Of course, details will depend on the specifics of the building but these are typical examples of areas which should be considered : Public entrance and lobby Loading dock Spaces occupied by other building tenants ADP facility reception area ADP input/output counter area 45 [PDF page 48] FIPS PUB 31 ADP data conversion area Tape library Systems analysis and programming areas Computer rooms Communications equipment rooms Air conditioning and other mechanical or electrical equipment spaces At this point it will be worthwhile to conduct a complete survey of the ADP facility and its environs to determine exposures, to verify se¬ curity measures already in place and to deter¬ mine from first hand inspection the state of current practice. GSA provides the following instructions for a physical security survey of a facility (exclusive of internal ADP areas) : 5.1.1. Instructions for the Facility Physical Security Survey A. Obtain a current floor plan which depicts all areas within the facility to include all access points and any adjacent areas belong¬ ing to the facility, such as parking lots and storage areas. B. Begin the survey at the perimeter of the facility and note the following: 1. Property line to include fencing, if any. and type. Condition, number of openings as to type and use, and how secured. Are there any manned posts at the property line, 2. Outside parking facilities. Is this area enclosed and are there any controls? Is the parking lot controlled by manned posts or are devices used? 3. Perimeter of facility. Note all vehicular and pedestrian entrances and what con¬ trols are used, if any. Check all doors— number, how secured, any controls or de¬ vices, such as alarms or key card devices. Check for all ground floor or basement windows—how secured ; screening, bars, etc., and vulnerability. Check for other entrances such as vents, manholes, etc. Are they secured and how? Check for fire escapes—number and location and acces¬ sibility to interior of facility from fire escape (windows, doors, roof). How are accessways secured? 4. Internal security. Begin at the top floor or in the basement. Check for fire alarm systems and devices noting the type, loca¬ tion, and number. Where does the alarm annunciate? Check telephone and elec¬ trical closets to see if they are locked. Are mechanical and electrical rooms locked or secured? Note any existing alarms as to type and number. Where do the alarms annunciate? Determine num¬ ber and location of manned posts, hours, and shifts. 5. Monitoring facility. Location, who moni¬ tors, who responds, type, and number of alarms being monitored. C. The following questions should also be in¬ cluded in a physical security survey: 1. Is the installation/building protected by alarm system (s) ? 2. How many zones of protection are within the protected building? 3. Is the alarm system adequate and does it provide the level of protection required? 4. Are there any vulnerable areas, perime¬ ter, or openings not covered by an alarm system ? 5. Is there a particular system that has a high nuisance alarm rate? 6. Is the alarm system inspected and tested occasionally to insure operation? 7. Is the system backed up by properly trained, alert protection officers who know what steps to take in case of an alarm? 8. Is the alarm system regularly inspected for physical and mechanical deteriora¬ tion? 9. Does the system have tamper-proof switches to protect its integrity? 10. Do system (s) have environmental or pro¬ tective housing or covers? 11. Is there an alternate or separate source of power available for use on the system in the event of external power failure? 12. Where is the annunciating unit located— local, central station, etc.? 13. Who maintains the equipment and how is it maintained (contract, lease equipment, force account personnel) ? 14. Is the present equipment outdated? 15. Are records kept of all alarm signals re¬ ceived to include time, date, location, action taken, and cause of alarm? 46 [PDF page 49] FIPS PUB 31 16. Are alarms generated occasionally to de¬ termine the sensitivity and the capabili¬ ties of systems? When the physical security survey is com¬ pleted, it should provide a picture of the exist¬ ing alarm systems and their location and also the number and location of manned posts, the number of personnel at these posts, and their schedule. With these facts in hand, the ADP security planner can proceed to the evaluation of exist¬ ing access controls and protection measures, identification of areas where remedial measures are needed and selection of specific measures. The sections which follow describe a variety of useful controls and measures which are in¬ cluded here for general guidance. However, one should seek help from the building manager and the Federal Protective Service (FPS) of the General Services Administration. To the extent permitted by the availability of person¬ nel, the FPS will perform a building security survey on request and can also provide expert advice and guidance on security hardware and the services which can be provided by Federal Protective Officers or contract guards. The use of various types of devices to aug¬ ment the existing protective force should be considered. Through the use of such devices, it may be possible to eliminate some of the sta¬ tionary manned posts at both vehicle and and pedestrian entrances. The manpower thus freed could be directed to other areas or fa¬ cilities. 5.2. Boundary Protection The threat analysis may indicate the need to protect the boundary of the property on which the building is located. This may be done by installing fences or other physical barriers, outside lighting, perimeter intrusion detectors or by using a patrol force. Often a combination of one or more of these will be effective. Fenc¬ ing may be high enough to deter the casual trespasser (three or four feet), too high to climb easily (six to seven feet) or may be intended to deter the determined intruder (eight feet high with three strands of barbed wire). In some cases it may not be necessary to fence the entire area. One may concentrate on key areas such as truck dock areas, parking areas (particularly for nighttime use) or por¬ tions of the building which are difficult to keep under surveillance. Alternatively, one can consider the use of extensive lighting to discourage prowlers. This may be the preferred solution where the threat level is low and fencing is not desired for cost or appearance reasons. Critical areas, en¬ trances, parking areas and locations not cov¬ ered by existing street lights should receive special attention. In those situations where an entrance is protected by a guard stationed in¬ side, or is used by personnel exiting after dark, it is wise to provide ample exterior lighting. Likewise it is advisable to avoid the use of tinted glass in such locations, as it may be difficult or impossible to see outside after dark. A third technique for perimeter protection is to use detection devices, usually infrared or microwave beams, which will be interrupted by an intruder. Such devices will cost in the range of $1 to $7 per linear meter and avoid the unsightly appearance of a fence. However, they are not as effective in deterring trespass¬ ers, have no value for crowd control, and probably can be circumvented by the skilled intruder. Furthermore, if intrusion detectors are to be useful, one must provide for prompt and effective response by guards when there is an alarm. Depending on the characteristics of the device used and the locale, one must expect false alarms as well. For all these rea¬ sons intrusion detectors are of limited value except as a back-up to fencing where a high level of perimeter protection is required or in certain special circumstances where fences are not feasible. In situations where one is concerned about intruders climbing over or slipping under a fence, one can equip the fence with vibration sensors. One such system uses small sensors mounted on every second or third post and at each gate. Sensors are connected by a continu¬ ous wire run to a control panel. Fence motion equivalent to an effort to climb the fence will cause an alarm. The cost is in the range of $1 to $3 per linear meter of fence. When the ADP facility building is part of a group of Federal buildings and the threat level is judged to be high and fencing is not prac¬ tical, an outside patrol force may prove to be the most effective protective measure. The com¬ position of the patrol force, its resources (ve¬ hicles, radios, dogs, etc.) and standing orders should be carefuly worked out to meet protec¬ tion needs at least cost. As a rule these decisions will be made by the FPS. The ADP security planner will want to understand the level of protection being provided, and be satis¬ fied that it is adequate to meet the needs of the ADP facility or, if necessary, seek appropriate adjustments. In some situations, for example, an emnloyee parking lot in a high crime area, it may be helpful to provide a low light-level, closed-cir¬ cuit television (CCTV) system for nighttime surveillance. Such a system uses one or more CCTV cameras located to cover the desired area and connected to monitors at a central security location. Typically each camera will be on a pan-tilt mount and have a zoom lens, both of which can be conti’olled from the monitor. 47 [PDF page 50] FIPS PUB 31 These features will permit the operator to watch a wide area for general activity or to zero in on a particular spot. Depending on in¬ stallation and specific features, each camera- monitor pair will cost from $4,000 to $10,000 or more. Hardware should be specified by a properly qualified and experienced person. It should be understood that it is unrealistic to expect the operator to watch the monitors alertly for long time periods. Either he should have a schedule for periodic sweeps, or intru¬ sion detectors should be provided to alert him to unusual events. However, a well planned and properly used CCTV system can permit a single guard to monitor a wide area often at a lower cost than a roving patrol. An exterior CCTV surveillance system can also be of great value for a facility which is subject to demonstrations or other crowd con¬ trol requirements. Because he can see the entire situation at a glance, the security director can control his security forces in “real time” to assure that the appropriate level of force is applied at all times and to respond promptly to changing conditions. This technique has been used with great success at a major Federal research facility. It should be noted that prior to the procure¬ ment of CCTV equipment for use in GSA oper¬ ated buildings, proposals must be submitted to the Office of Federal Protective Service Man¬ agement, Systems Branch, for concurrence. To summarize briefly: ® Fences or other barriers will provide crowd control, deter casual trespassers and help in controlling access to entrances, but it can be costly, will not stop the determined intruder and may be unacceptably unsightly. ® Intrusion detectors can alert a guard force to intruders and may be practical where a fence cannot be installed, but they are subject to nuisance alarms, can probably be penetrated by the skilled intruder and require human re¬ sponse to alarms. ® A patrol force can provide flexible response (particularly in emergencies), and good deter¬ rence and may be particularly effective for protection of a group of buildings. However, the cost may be excessive. ® CCTV systems permit one man to monitor a large area and see exactly what is happening but should be coupled with an alerting function (intrusion detectors or scheduled scanning) and the provision for human response. 5.2.1. Emanations In evaluating the need for perimeter protec¬ tion, the ADP security planner should take into account the possibility that electromagnetic or acoustic emanations from ADP hardware may be intercepted. Tests have shown that inter¬ ception and interpretation of such emanations may be possible under the right conditions by technically qualified persons using generally available hardware. As a rule of thumb, inter¬ ception of electromagnetic emanations beyond 300 meters is very difficult. However, if the ADP security planner has reason to believe that there may be a potential exposure to inter¬ ception he should seek technical guidance from qualified vendor representatives. The choice be¬ tween physical separation of radiating devices from potential intercept points and the use of screening should be based on an analysis of relative cost. Particular attention should be paid to remote terminals which may be located in commercial buildings with non-government tenants. 5.3 Entrance Door Controls The objective of perimeter protection is to deter trespassing and to funnel employees, visi¬ tors and the public to selected entrances. The objective of entrance door controls is to screen entrants, to deny entrance where appropriate and to control the flow of materials into and out of the building. Screening can be done in two ways: personal recognition of the entrant or acceptance of credentials by a guard* or by the possession by the entrant of a suitable device to unlock the door. Screening by a guard is by far the most positive when applied conscientiously but will cost in the range of $2 to $10 per hour per entrance depending on circumstances. Entrant screening can be accomplished by electronic or mechanical devices. Authorized entrants may use a key (conventional or electronic), enter the combination of a push button lock, or be screened by a device which compares an en¬ trant characteristic (hand geometry, finger¬ print or voice characteristics) with stored information about authorized entrants. Access control which depends on a key lock or screen¬ ing device in place of a guard suffers from several shortcomings. Keys or combinations can fall into the wrong hands. An intruder may enter immediately behind an authorized en¬ trant (often referred to as “tail gating”). The skilled intruder may defeat the lock. While these shortcomings can be managed (careful key control, security conscious employees, burglar-alarmed doors, etc.), the ADP security planner should be aware of these problems and not fall into the trap of accepting blanket state¬ ments like “This door is always locked”, or “This key cannot be duplicated.” The features of various door control devices can be sum¬ marized briefly as follows: * Reference to a guard for screening entrants should be taken to mean any person who has specific screening responsibility and thus may include a receptionist, truck dock supervisor or clerk at a computer in place of a uniformed security guard. 48 [PDF page 51] FIPS PUB 31 Conventional keys and lock sets. Cost is minimum, less than $1 per key and about $5 per cylinder. Almost any door type can be equipped. However, keys are easily duplicated and locks can be picked. A key holder can enter at any time. There is no control over en¬ trance and exit of materials. Pick resistant lock sets. Cost is about two or three times higher than conventional locks, keys are much more difficult to duplicate and locks are much harder to pick. Other character¬ istics are the same as conventional locks. Electronic key system. These use specially encoded cards to actuate an electric door strike. (With a conventional lock set, the key is used to withdraw the bolt from the strike, thus permitting the door to open. With an electric strike, the bolt remains extended and an elec¬ tric solenoid retracts the door strike to allow the door to open.) Depending on features and installation, cost will range from about $400 per door to several thousand dollars per door. Cards may cost several dollars each. Simple systems perform as pick-resistant lock sets. At higher cost the system can include the ability to lock out specified cards, to limit access to speci¬ fied times, to log all entrances and exits, and to control a group of doors such that access to each door in the group can be specified for each card. Electronic combination locks. Such locks typi¬ cally have electronic push buttons into which the entrant keys the combination to actuate an electric strike. Costs and features are generally similar to electronic key systems except the entrant need not carry a card. Some allow the entrant to use a special code when under duress which will open the door but at the same time sound a remote alarm. Cost is about $500 per door. Mechanical push button combination locks. Pressing the correct combination allows one to retract the bolt and open the door. The special features described for electronic locks are not available, but the cost is much lower, typically $40 to $80 per door. Physical characteristic locks. Cost is in the range of thousands of dollars per door and may require the entrant to carry an electronic key card. These systems come the closest to duplicating human screening in that they meas¬ ure some physical characteristic of the entrant such as hand geometry, a fingerprint, etc. How¬ ever the accept-reject decision is made on the basis of an analog input and so some errors will be made, i.e. entry will be denied to an authorized entrant, and vice versa. Further¬ more, since such devices are relatively new, it is not yet clear how reliable they are and how easy it may be to circumvent them. If it is determined that personal screening is necessary at a number of doors and traffic at each is relatively light, it may be cost effective to have a single guard control these entrances with a closed circuit TV (CCTV) system. Each door is equipped wtih a TV camera, a signal¬ ing device, an intercom and an electric door strike. To control both entrance and exit it is necessary to have two controlled doors with a vestibule between. This may lead to conflict with emergency exit requirements so caution in planning the installation is required. One com¬ mercially-offered system includes a special TV camera which presents a close-up view of the entrant’s photo-identification card. By also viewing the entrant on the CCTV monitor and talking to him on the intercom, the guard can screen the entrant almost as effectively as he could in person. Note that he can also monitor movement of materials. The cost for hardware will be in the range of $3,000 to $6,000 per entrance but will be quickly recovered in sav¬ ings in labor. Since the screening may permit only four or five entrants per minute, one should analyze the traffic patterns carefully, particularly at shift changes, to be sure that there will be no undue delays. Such delay of personnel on an hourly payroll could lead to added expense, a point which should be consid¬ ered for any unusual screening technique, in¬ cluding CCTV. It can be seen that at gradually increasing cost one can impose ever more effective screen¬ ing of personnel and materials. Every effort should be made to establish requirements care¬ fully for each entrance to avoid needless ex¬ pense and unnecessary entrances should be eliminated if possible. Each entrance door should be capable of re¬ sisting forced or covert entry up to the level of effort which is likely to be applied. This entails careful consideration of door hardware and installation. Where appropriate one may use heavy-duty lock sets, reinforced strike plates and door frames, tamper-resistant hinges and break-resistant glass in vision pan¬ els. The ADP security planner should seek advice from qualified persons in this area. In addition to reinforcing doors one may also connect critical doors to a perimeter alarm sys¬ tem to signal a guard when a door is opened. This can be done for electric strike-equipped doors in such a way that an alarm is not sounded when normal entry is made but forced entry will cause an alarm. 5.4. Perimeter Intrusion Controls One should check the perimeter of the build¬ ing for other possible entry points such as win¬ dows, transformer vaults, air conditioning louvers, roof hatches and the like. Each point which represents a potential intruder route should be appropriately secured physically or added to the perimeter alarm system. For example, exposed windows can be glazed with break-resistant glass or plastic. Louvers can 49 [PDF page 52] FIPS PUB 31 be protected with heavy gauge screens. The determined intruder may even break through a wall or roof if he feels he will be unobserved for a long enough time period and the target is worth the effort. Where physical protection or adequate surveillance against such forced entry is not practical (as, for example, in a building not controlled by the government) one may install special sensors at windows, loading docks or around the entire perimeter of the building if needed. The electromechanical type of intrusion de¬ tection system is in widest use today. It con¬ sists of a continuous electrical circuit so bal¬ anced that a change or break in the circuit will set off an alarm. Some examples of systems using a continuous electrical circuit are: foil strips on a window that will break if the win¬ dow is broken, magnetic or contact switches on the doors, mercury switches on openings that tilt, vibration detectors to detect breaking through walls, and screens and traps which consist of fine wires imbedded in breakable dowels or in the walls, ceilings, and floors. Any tampering with the mechanical parts of the system or breaking or grounding of the elec¬ trical circuitry will cause an alarm in the cen¬ tral station. These devices are relatively simple and are normally used for perimeter protection. They may be added to any system (local, pro¬ prietary, etc.) without interfering with other detection devices. The various kinds are listed below: Window foil. Window foil is a metallic tape affixed to windows and glass doors. When the glass is broken, the foil breaks, an open circuit results, and an alarm is sounded. A hairline crack or scratch will activate the system caus¬ ing an alarm. Wire lacing and screening. This electrome¬ chanical device uses fine wires laced across door panels, floors, walls, and ceilings. A forced entry into the protected area will break a strand of the laced wire which will cause an alarm. Taut wire. A taut wire device is used to detect intrusion into a protected area. A fine strand of wire is strung under tension across internal openings such as air ducts or utilities tunnels. Any change in the tension of the wire will cause an alarm. Intrusion switch. A magnet or mechanical intrusion switch is frequently used to protect doors, windows, skylights, and other accessible openings. Switches may be surface mounted or recessed. • Magnetic intrusion switch. This switch consists of two parts, one being the mag¬ net, the other a switch assembly. When the magnet is properly oriented and mounted adjacent to the switch assembly, the switch is activated. When it is removed the switch is deactivated and an alarm is sounded. Usually the magnet is mounted on the movable portion of the door, win¬ dow, or item protected. • Mechanical intrusion switch. This switch is also activated by opening a door, win¬ dow, skylight, etc. The plunger type switch is usually recessed and costly to install. The lever type switch is less expensive to install but is easily detected. Mechanical switches exposed to the weather may stick or freeze. In summary, entry into a building is best controlled through either surveillance or high integrity access controls at desired points of entry and by either surveillance or alarm sys¬ tems around the remainder of the building perimeter. A recent report, “Penetration Tests on J-SIIDS Barriers” [21] shows very graphi¬ cally how inadequate most structures are for stopping a determined intruder. The report describes actual tests of the time required to make an 8” x 12” opening in a wall, the size judged to be the minimum required by an in¬ truder. Results can be summarized briefly as follows: Wall Construction Tools Used Penetration Time 2" x 4" studs with 1" siding both sides Hand brace and electric sabre saw 1.55 minutes 8" cinder block wall Sledgehammer 1.52 minutes* 8" cinder block wall with brick veneer on one side Sledgehammer 2.12 minutes* 5-14" reinforced concrete Rotohammer drill and sledge¬ hammer 5.44 minutes* 8" reinforced concrete Rotohammer drill and sledge¬ hammer 10 minutes approx.* * Add approximately 1 minute for each reinforcing rod encountered. 5.5. Critical Area Controls Within the ADP facility, there may not be equal access to all areas even when it is as¬ sumed that everyone in the building has been screened through the building perimeter controls. The following areas constitute a minimum set to be analyzed to determine per¬ missible access, both during operational periods and when the facility is closed: Computer room Data storage library Input/output area Data conversion area Programmer areas/ files Document library Communications equip¬ ment area Computer maintenance room Mechanical equipment room Telephone closet Supplies storage 50 [PDF page 53] FIPS PUB 31 In addition to protecting the confidentiality and integrity of data files, areas should be con¬ sidered with regard to protecting valuable as¬ sets, preventing tampering, vandalism and sabotage, and preventing the perception of op¬ portunities for malice and mischief through unauthorized browsing. The objective of the analysis is to identify all sensitive or critical areas and determine from a study of work flow and job assignments which persons are to be given access and at what times. The next step is to select control methods. The basic techniques which apply to exterior doors apply here but with two signifi¬ cant differences. First, it is expected that such areas would be either unoccupied and locked or occupied by authorized personnel. If clear regulations have been published and affected persons properly briefed as described in Chapter 9, then it is reasonable to expect unauthorized persons to be challenged if they enter the space while it is occupied. Second is the important requirement to avoid impeding work flow unnecessarily. This means that the ADP security planner should examine work flow, people, information and materials carefully in relationship to the physical layout of the ADP facility to avoid obvious problems, such as placing a secured area in the path between two less critical areas. Furthermore, one should try to avoid situations where the designated access route to a controlled area is circuitous and a shorter but unauthorized route (e.g., a fire exit) is available. In such cases there will be a natural tendency to use the short cut. But even when the designated route is convenient, it is not uncommon to find fire exits misused. The common solution for this is to place alarm actuators on fire exit doors. If the facility has a central alarm sys¬ tem, a signal should go to the central system whenever a fire exit door is opened. However, for maximum effectiveness, the alarm should be audible at the doorway. There are self con¬ tained alarm boxes which may be mounted on fire doors. The typical alarm is about 10 x 20 x 7 cm in size and has a key actuated arm/ disarm switch. When the door is opened, a loud alarm, powered by an internal battery, comes on and can continue to sound until turned off with the key. The cost is approximately $60 per alarm. The ADP security planner should remember that efforts to control access must not conflict with life safety objectives. The NFPA “Life Safety Code” [30] defines the number, size, and location of fire exits as a function of the building occupancy and construction. It is im¬ portant to see that there is compliance with such standards and with applicable Federal regulations. There are several technological means of de¬ termining access to or occupancy of critical areas during periods when the areas should be vacant. Two have been discussed: light beams across entrances and CCTV systems. An im¬ portant caution is that CCTV systems are best used only for a determination of an area’s status after there has been an alert from some other, more positive intrusion detector. There are at least four distinct technologies for de¬ tecting the presence of an intruder: 1. Photometric Systems. These are passive systems which detect a change in the level of light in an area, due to added sources of light, or reflections or absorptions of existing light. Since these systems are sensitive to ambient light levels, they may be used only in windowless areas (or areas in which the windows have been covered). 2. Motion Detection Systems. The basis for the operation of these systems is the Doppler effect. When the source of a sound or electromagnetic signal, or a re¬ flector of such a signal, moves toward or away from a receiver, the frequency or pitch of the signal received will be higher or lower, respectively. In a room having a source of wave energy and a receiver, if a body moves within that room, the motion can be detected from the change in frequency of received wave¬ forms. The receiver will pick up the source frequency strongly, but will also detect a slightly different frequency at a much lesser strength. • Sonic. Sonic detection systems oper¬ ate in the audible range, 1500 to 2000 hertz and higher. The constant tone is very annoying since it is well within the audio range and at a high decibel (DB) output. This system uses trans¬ mitters and receivers (transducers) to saturate the entire enclosure with sound waves. These transmitting and receiving transducers are permanent magnet (PM) speakers and are mounted within the same room, usu¬ ally on walls opposite each other. The receiver listens to the tone being transmitted and compares the reflected signal. Whenever the pattern of the tone varies due to a disturbance with¬ in the protected area, the receiver detects this change in frequency and activates an alarm. 51 [PDF page 54] FIPS PUB 31 • Ultrasonic. The ultrasonic detection system utilizes high frequency sound waves with a frequency of about 19000-20000 hertz, but are otherwise like the sonic systems. Since the fre¬ quencies used are at the upper limit of the audible range, only a few per¬ sons (generally children) can hear them. • Microwave. The microwave system operates in a similar manner to the above systems. The difference is that microwaves are high frequency radio waves. These radio waves are trans¬ mitted at a frequency between 400- 10,000 megahertz. Microwave signals can be controlled as to the size of the area to be protected through selection of the type of antenna used. One or several antennas can be used in a given location. Single or multiple units can be used to provide the re¬ quired protection without interfering with sonic or ultrasonic units. 3. The Acoustical-seismic Systems (audio). This system employs microphone-type de¬ vices to detect sounds which exceed the ambient noise level of the area under pro¬ tection. It is obvious that they cannot be employed in areas where noise from man-made sources, such as aircraft, con¬ struction, etc., are likely to set off nuisance alarms. Some are even triggered into alarm by the elements, such as rain or thunder. Some acoustic systems rely upon air to transmit the sound to the microphone-type device. Others will not respond to ordinary noises in the air but only to those transmitted through a structure such as a wall. • Acoustical (audio). An audio detec¬ tion system listens for intrusion sounds by using microphones installed in the protected area. Upon detection of intrusion sounds, an alarm occurs. This type of system may be equipped with cancellation and discrimination units which electronically evaluate the significance of the sound disturbance, thus eliminating reaction to nuisance alarms which may be caused by air¬ planes, thunder, etc. • Vibration (seismic). This type of sys¬ tem utilizes the same principle as the audio detection system except that highly sensitive and specialized micro¬ phones are attached directly to objects such as safes, filing cabinets, windows, walls, and ceilings. Vibration of these objects initiates alarms. Cancellation and discrimination units are neces¬ sary to prevent nuisance alarms. 4. Proximity Systems. There are various types of proximity systems all of which detect the approach or presence of an object or an individual. In principle, a proximity system employs an electrical field which, when upset by a foreign body, causes an alarm. The field may be set up around a cabinet or it may simply surround a wire. Whether the field is electromagnetic or electrostatic, the prin¬ ciple of balance and unbalance applies. There are several methods of establishing the field; methods differ to some extent among manufacturers. A proximity sys¬ tem may also be employed to protect an area by erecting what is commonly known as a magnetic fence; that is an integral part of the system. Other variations pro¬ vide surveillance of doors and windows. The proximity system is designed to be supplemental and cannot be used effec¬ tively as a primary system. This is be¬ cause of the system’s susceptibility to nuisance alarms caused by electric supply fluctuations and by the presence of mops, pails, etc., placed near the system. Ani¬ mals and birds can trigger a system into alarm if it is too sensitive. Therefore, proximity systems should be backed up by other security systems. The following table compares six of the more commonly available interior surveillance sys¬ tems : Approx Sensor Type Cost Limitations Resistance to Defeat Photometric $500 Extraneous light must be excluded from area; limited to interior rooms. High Motion Ultra- Sonic $250 Air motion may cause false alarms. Moderate to High Motion micro- wave $500 Energy can penetrate walls, etc. causing nuisance alarms. High Acoustical- seismic, sound $250 Extraneous noises will generate nuisance alarms. High Acoustical- seismic, vibration $100 Localizing the source of nuisance alarms could be difficult. High Proximity. capacitive $350 Susceptible to nuisance alarms; require backup. High In planning the security for critical areas one may make use of the intrusion detectors already described, the controls which can be imposed by guards or personnel assigned to 52 [PDF page 55] FIPS PUB 31 the areas or the physical barriers created by internal partitions. In the latter case, the ADP security planner should check construction de¬ tails carefully. In modern office buildings using hung ceilings, interior partitions may not ex¬ tend above the ceiling. This means that an intruder may be able to enter a room by lifting a ceiling panel and climbing over the partition; this is a particularly troublesome form of in¬ trusion since it can be done quickly and quietly without tools and will leave no sign of forced entry. Likewise, interior partition door frames are often of lightweight construction and easily forced open. The key point is not to place undue reliance on interior partitions. 5.6. Guard Force Operations Physical protection measures, physical bar¬ riers and intrusion detectors depend ultimately on human intervention. Where there is a need for full time guards, they will either be Federal Protective Officers provided by the Federal Protective Service of GSA or guards furnished by a private company under contract. In as¬ sessing the role guards can play in supporting the ADP security program, it is helpful to review the kinds of tasks which can be as¬ signed to them. First a guard may be assigned to a fixed post: a lobby, entrance door, truck dock, en¬ trance gate or security control desk. His post orders may include: • Checking entrant credentials and use of the sign-in log. • Issuing and recovering visitor badges. • Monitoring intrusion and fire alarm sys¬ tems and dispatch personnel to respond to alarms. • Controlling movement of materials into and out of the building and enforcing property pass regulations. • Enforcing rules and regulations estab¬ lished for the building. • Accepting registered mail. To make optimum use of a guard it is im¬ portant to see that his post orders are complete and clear and that he is properly trained. For example, if the guard is to control the move¬ ment of tapes, disks and other ADP media, he must be able to recognize them and understand what they are. If a guard must devote his time and attention to receiving visitors, preparing badges and telephoning for escorts, he cannot be expected to check employee credentials vig¬ orously at the same time. The ADP security planner who intends to make use of a specific guard post to support the ADP security pro¬ gram, should review the guard’s post orders and work load with the building security direc¬ tor to be sure his expectations can be met. Second, a guard may be a roving patrol guard with a specific route or a general area which he may cover on foot or in a vehicle. His duties may include these functions: • Verify that doors, windows and other openings are properly locked during des¬ ignated periods. ® Observe and correct or report safety haz¬ ards such as immediate fire hazards, equipment or machinery left on, stumble hazards, fire doors propped open and the like. • Verify the condition of fire extinguishers, hose lines and automatic sprinkler sys¬ tems. • Check that files, safes and restricted areas are properly secured. • Be alert to suspicious persons or activity, unusual odors, leaks or other abnormal conditions. If he is to be effective, the roving guard must be under some kind of control. This means either that he reports to a control point at regular intervals either in person or by tele¬ phone, or that he is provided with a portable two-way radio. In the latter case he can be dis¬ patched to the scene immediately should an emergency arise. As with the fixed post guard, it is important for the ADP security planner to see that the roving guard has the necessary orders and training to protect the ADP facility properly. For example, if the roving guard smells smoke in an unattended computer room, what should he do beyond giving the alarm? Can he turn off electric power and, if so, does he know where the disconnect switch is lo¬ cated? Similar questions about air condition¬ ing, plumbing leaks and other ADP related emergencies during unattended hours should be analyzed carefully and appropriate orders formulated and guards trained to carry them out. There is a final point which should be con¬ sidered when developing the security indoctrin¬ ation program described in Chapter 9. There is often a tendency for professional staff mem¬ bers to think of the Federal Protective Officer or private contract guard as unimportant and unworthy of consideration. Apart from human feelings, this attitude can nullify the contribu¬ tion which the guard is depended upon to make to ADP security. ADP management and senior staff members should, by willing compliance with regulations and their general behavior, display their support for the guard in carrying out his assigned duties. 53 [PDF page 56] FIPS PUB 31 5.7. Integrating Physical Security Measures The preceding sections of this chapter have discussed the various techniques for providing physical protection. It is not uncommon to find that as each new security or emergency re¬ sponse requirement is discovered (often as the result of a specific event) at an ADP facility, some countermeasures are taken to deal with it. As a result the overall physical protection program evolves piecemeal and so is usually un¬ even, expensive and cumbersome. On the other hand, a careful examination of the totality of security and emergency requirements, people and procedures will often show how they can be integrated for maximum effectiveness at least cost. For example, these guidelines have discussed the following kinds of security hardware systems: ADP area smoke detection systems Sprinkler system flow alarms Building-wide fire alarm pull-boxes Perimeter intrusion detectors Door status detectors Critical area intrusion detectors Area surveillance CCTV Entrance control CCTV Electronic door locks As required by particular circumstances, the physical protection plan may use several of these systems. While one may specify and pro¬ cure each needed system separately, planning for all requirements as an integrated whole can have two major benefits. First is the require¬ ment for human response to each alarm con¬ dition. Consolidating alarm control panels and CCTV monitors in the least number of loca¬ tions will minimize the number of people re¬ quired to do this. Second, one may find that more sophisticated alarm controls can be used. One approach uses multiplexor techniques to connect many alarm points to a single control unit via a single circut with substantial sav¬ ings in wiring cost and improved maintainabil¬ ity. Typically more than one sensor type can be connected to the individual alarm points. More advanced systems use a process-control mini-computer to control electronic access doors, monitor alarm sensors and building me¬ chanical equipment. In addition to integrating hardware, the ADP security planner, working with the build¬ ing manager and building security director, should consider the human resources available to support the physical protection plan. In addi¬ tion to full-time guards, the following people may, as permitted by regular duties, be able to participate: Receptionists and information desk per¬ sonnel Building engineering staff Building and grounds maintenance staff Shipping and receiving clerks Messengers Area supervisors Mail room personnel By considering where such people are located and the needs of the physical protection plan, it may prove possible to get the needed re¬ sponse to alarm situations with a minimum number of guards. However, it can be seen that to do so, thought must be given to the location of security systems, particularly alarm indicators. We have purposely omitted from this chap¬ ter detailed information on security hardware and alarm systems for two reasons. The tech¬ nology is developing rapidly and new devices appear on the market almost daily. In addi¬ tion, the Federal Protective Service of GSA can be called upon for detailed advice and ex¬ pert guidance in meeting specific requirements. When physical protection plans have been completed, the ADP security planner should check two final points. First, great care should be taken to see that plans and specifications for the ADP facility and its security hardware, alarms, locking systems and related items are protected against disclosure except on a need- to-know basis. Second, the emergency response plans and physical protection measures should be carefully integrated to assure maintenance of security during an emergency. For example, one must guard against the use of a nuisance fire alarm and the resulting evacuation to cir¬ cumvent controls over access to key areas. 60 Interna! Controls 6.0. Introduction The four preceding chapters have presented physical means for supporting ADP security objectives. This chapter discusses the use of internal controls to reinforce physical safe¬ guards in four areas: personnel, organization structure, the data base and programming. Generally speaking it will not be necessary to cost justify internal controls solely on the basis of expected loss reduction since controls will usually be installed to serve other objectives as well, e.g., cost accounting, error detection and correction, management reports. It is 54 [PDF page 57] FIPS PUB 31 likely that the ADP security planner will find that needed controls already exist and that his task will be to determine what modifications and extensions are needed. The basic risk an¬ alysis will have identified sensitive areas and applications. Physical security measures will require human intervention, support and co¬ operation. The ADP security planner should bear these factors in mind as he reviews the sections which follow to be sure that internal controls are structured to reflect security ob¬ jectives. 6.1. Personnel Controls People are undoubtedly the most important part of the ADP facility, and no ADP facility can function without a trained staff dedicated to achieving the mission of the agency. Per¬ sonnel controls should reflect the need for care¬ ful selection of mature, trustworthy people for sensitive positions, the importance of provid¬ ing adequate training to assume competent per¬ formance of ADP duties, and the value of good supervision in achieving a high level of moti¬ vation. 6.1.1. Personnel Selection The selection of personnel routinely in¬ cludes an effort to determine that the candi¬ date is qualified by training, talent and ex¬ perience to perform the duties to be assigned. In addition to this determination of job skills, the selection process for sensitive ADP posi¬ tions should also verify the trustworthiness of the candidate for sensitive positions by appro¬ priate pre-hire screening. Several levels of screening are available and, of course, both ef¬ fectiveness and cost increase as the depth of the investigation increases. Therefore the level of screening used should reflect the relative sensi¬ tivity of each position. Each ADP facility must define for itself its sensitive positions; gener¬ ally these will include computer operations, data control, management, auditing, and pro¬ gramming (including acceptance testing and maintenance) of critical applications and sys¬ tems. The risk analysis for fraud will usually identify critical interface points. Wherever a critical interface involves a single indivdual, the postion is probably sensitive. This is espe- cally true for hidden interfaces in which checks and balances are missing, e.g., a single pro¬ grammer has the responsibility for creating, testing, debugging, and installing a critical pro¬ gram. The most sensitive position is often that of the system programmer; a qualified prac¬ titioner of operating system maintenance can do more damage with less chance of being caught than almost any other person involved with data processing. Each Federal Department or independent agency has established regulations and pro¬ cedures for designating one or more levels of position sensitivity and the screening applied to each sensitivity level. The ADP security planner should establish the appropriate level to apply to each ADP facility position. Person¬ nel procedures should be established to insure that Item E, Position Sensitivity of Part I of U.S. Civil Service Commission Form 2—Re¬ quest for Personnel Action, properly reflects the sensitivity levels selected. 6.1.2. Training A surprising number of operations problems and security breaches result from promoting an individual into a position beyond his compe¬ tence. Rather than admit defeat, such people have been known to destroy source documents or falsify reports in an attempt to conceal shortcomings. The ADP facility can use its personnel train¬ ing program to minimize such security and in¬ tegrity problems. The training for each specific job should be thorough, efficient, and compe¬ tent. But strong motivations is just as essen¬ tial as technical competence. Each employee should be given an adequate orientation to the agency, its mission, the ADP facility and his own career development opportunities. Person¬ alized security training is essential. It should include not only the objectives of the security program and its operation but the duties and obligations of each staff member as well. De¬ tails are given in Chapter 9. 6.1.3. Supervision Each ADP supervisor can make a strong contribution to the security program in several ways. First, he can see that he and his staff comply with both the letter and the spirit of security regulations and control procedures. He can also actively seek out effective ways to to improve security. Next, the good supervisor will work at main¬ taining close, effective communications with his staff. He should try to be sensitive to feelings and attitudes so that he can act affirmatively in cases of potential disgruntlement. It is much better to seek resolution of conflict situations than to ignore them, as unresolved conflict can only lead to frustration and impulsive action. Finally, the good supervisor will take pains to see that each member of his staff is compe¬ tent in his assigned duties. While incompetence cannot be tolerated in any work situation, the consequences can be particularly pervasive in an ADP facility. A program will faithfully re¬ peat an erroneous instruction indefinitely. A moment of careless operation can damage hard¬ ware or destroy a file. Staging the wrong tapes 55 [PDF page 58] FIPS PUB 31 can delay jobs. While errors and lapses can never be completely eliminated, the conscien¬ tious supervisor will do his best to match the individual to the job and to give him needed support and training. 6.2. Organizing for Internal Control One of the basic principles of internal control is to divide the execution of critical functions between two or more persons, a technique often referred to as separation of duties. The theory is that errors are less likely to go undetected when several people review the same trans¬ actions and fraud is deterred if there is a need for collusion. One individual should never be totally responsible for a given activity espe¬ cially if it relates to the processing or develop¬ ment of sensitive applications. This principle of two individuals acting in concert, yet independ¬ ently, to effect action can be applied to data processing operations. The best approach to determine the exact points where separation of duties should occur is to identify the loss tar¬ gets by referring to the basic risk analysis for the ADP facility and then to identify the routes to those targets which an intruder could use. Finally, the points along the route can be iden¬ tified where separation of duties would provide a desirable level of protection. As a rule, sep¬ aration of duties will be required to control sensitive applications, to prevent compromise of access controls and to avoid abuses in the area of reject and exception processing. Figure 16 is a generalized diagram of a typi¬ cal ADP operation with potential control points indicated. The ADP security planner should review each sensitive ADP task to determine where controls would be effective in forestalling errors or fraud and determine how existing controls should be expanded to meet security needs. Consider payroll processing, for exam¬ ple: the controls should insure that input is accurate and valid and that output, paychecks, payroll journals, etc. do not fall into the wrong hands. If the payroll is large, exception proc¬ essing is probably important. Therefore, the clerk who prepares input should not control check signing and distribution or corrections to the payroll file. Similarly, the programmer who maintains the payroll program should not con¬ trol its acceptance testing. These examples are much simplified, of course. The real exposures are often hidden from direct view. The key point is to examine each potential target and identify the points in the work flow where separation of duties can help to stem losses. Many applications are designed for the rejec¬ tion of invalid input and its correction and re-entering. While this is a valuable quality control technique, the introduction of manual processing of rejects offers significant oppor¬ tunity for fraud as well as errors. A useful control for rejects processing is the use of a system-generated log or a bookkeeping journal record to keep track of all incompleted trans¬ actions. These records will provide an inde¬ pendent audit trail for control purposes, and separation of duties should apply to the clear¬ ing of the log. Someone other than the person responsible for correcting faulty input should initiate the transaction to clear log entries. Program and procedure change controls should receive special attention from the ADP security planner. The process of getting a pro¬ gram from test to production status exposes the system to compromise from unauthorized changes and to loss of data integrity caused by too hurried development or inadequate testing. The ideal approach to installing a change in a production program is a formalized system in which several different organizational functions are involved. User, programmer, auditor, and operations personnel should all be involved in the approval process. Quality control of pro¬ gramming is as important a concept as quality control in manufacturing. An organizationally discrete checking and follow-up function can be of value in maintaining program quality stand¬ ards. In addition, the larger ADP facility should consider establishing a separate testing func¬ tion for all programs that have reached final production status. Since controls are managed by people, the basic organizational structure must be respon¬ sive to the desired internal controls. Figure 17 shows a prototype organization chart. Note that the key control functions: testing and quality control, project management, input/out- put control, tape disk library and standards, security and data base administration have been separated from the production functions. This makes it easier to assure that checks and controls will function effectively. Of course, the details for a specific ADP facility will depend on its size and mission. While the major prob¬ lem for a large ADP facility is often effective control of resources, the major problem for the small ADP facility may be the practical prob¬ lems of having enough different people avail¬ able to implement desired separation of duties. If this is the case, and it is necessary for one or more individuals to have an unusually wide span of control, it may be necessary to depend on auditing. This presumes that good audit trails are provided. To summarize, the following points have been made: • Take great care in selecting personnel for sensitive ADP positions. Be sure that each per¬ son receives ample training and close, effective supervision. These measures will provide the basis for a strong ADP staff. • Analyze the tasks performed and assets controlled by the ADP facility to identify the targets and mechanisms for damaging errors or fraud. 56 [PDF page 59] FIPS PUB 31 Potential Control Points Figure 16. Work flow and control points. Dota Conversion Program Library System Design 57 [PDF page 60] FIPS PUB 31 Figure 17. Prototype organization chart. \ 58 [PDF page 61] FIPS PUB 31 • To the extent permitted by the size of the ADP staff use separation of duties at key con¬ trol points to minimize errors and deter fraud. • Augment separation of duties with in¬ ternal controls as appropriate to meet the se¬ curity objectives. 6.3. Data Controls Apart from conventional internal controls, the ADP security planner should particularly verify control and protection of data files. Care must be taken to see that information which has been designated as sensitive under Federal regulations is properly safeguarded when it is entered into ADP data files. This may require special handling, segregation or other tech¬ niques similar to those used for national se¬ curity information. The ADP security planner should also evalu¬ ate physical handling of data files at all points. He should examine the flow of data through the ADP facility to identify points at the input/ output interfaces, during handling, and during custodial storage, where controls may be needed to safeguard against possible loss or destruction —and equally important to assure that a loss will be detected. The ADP facility should follow defined procedures in case data is lost. Manual control techniques might include tape/disk movement control forms, inventory logs, au¬ thorization for use and special handling for critical items. The use of a computer system for control of data files deserves special consideration if there are a large number of files. Many vendor sup¬ plied tape or disk library management systems provide logging and control of tapes by volume, serial number and name; prevent unauthorized destruction of a data file; and provide auto¬ matic backup facilities. Such systems handle both on-line and off-line files. Similar systems are available to manage a program library. The typical system allows con¬ tinual modification of a program which is being- developed while retaining all previous versions. It protects against unauthorized modification, and helps with the management of program modifications. Such packages, whether pur¬ chased or developed in-house can be very useful for management and control of data and pro¬ gram files. In pre-computer days it was axiomatic to lock up sensitive or important information, ledger books and vital records in a desk drawer, file or safe when not in use. The same principle should also apply to valuable computerized data. The tape library should be locked when unoccupied and unauthorized persons should be excluded. Data safes and vaults, and data con¬ trol rooms should be protected in accordance with the sensitivity and value of the material (data) stored within. The exposure to magnetic fields should be evaluated realistically [12] and reasonable protective measures taken. Com¬ puter printouts should be destroyed in accord¬ ance with sound procedures to prevent disclo¬ sure. It does little good to develop extensive security controls against theft of data from the computer or programming area and then allow the same information to be available from waste baskets, loading docks or trash heaps. The ADP security planner should be sure that data control requirements are properly re¬ flected in the physical protection program described in Chapter 5. 6.4. Data Retention and Back-Up The preceding section has discussed protec¬ tion of current data files. The next step is to integrate the vital records management pro¬ gram with the data base management program to support common retention objecitves. Gener¬ ally speaking both short term and long term back-up is required. 6.4.1. Short Term Back-Up Short term back-up protects against localized or temporary loss such as cancellation of a job because of an interruption or error. The inter¬ ruption may last only a millisecond, and the program (especially if it is a short one) may be re-run easily. However, if the job is interrupted in the thirteenth hour of a fourteen hour proc¬ essing job, it would be wasteful to have to begin the job again. Therefore, checkpoints, restarting, recovering, and backup at inter¬ mediate points need to be considered for all long jobs. This is not news to anyone operating ADP facilities. Nonetheless, a consistent back¬ up program is rarely found. In determining short term back-up require¬ ments, cost considerations play a large role. For example, assume one could checkpoint at any time at a cost of X dollars. If the total job costs X dollars to run, it would not be cost effective to use any checkpoints. If it costs 200X to run the job, it would probably be sensible to back-up the data at intermediate points. A review of system reliability as de¬ scribed in Chapter 4 may be of help in making the best decisions. 6.4.2. Long Term Back-Up There are six reasons why one would want to retain a past environment: 1. Discovery of errors that caused data in¬ tegrity problems in the past, e.g. to trace a series of mistakes going back six months but not discovered until yesterday. 2. Back-up which permits disaster recovery. These situations are covered in detail in Chap¬ ter 8. 59 [PDF page 62] FIPS PUB 31 3. Management performance review or plan¬ ning. The future goals and activities of the ADP facility can be predicted more easily if information on past activities can be retained. Use of simulation models or other planning tools is enhanced if empirical data is used as input, 4. Statistical reporting requirements. Data from the past may be needed for analysis of trends and for extrapolations. 5. Audit requirements (internal and ex¬ ternal). The ability to analyze the past environ¬ ment is a primary requirement of the auditor. Specific requirements are discussed in Chapter 10. 6. Legal requirements. Other government agencies may need the data or there may be a statutory requirement to retain them. Any of these reasons would dictate that one should keep at least program source code, docu¬ mentation and data files which were in use at any given point in time. The ADP security planner should give thought to what is to be retained. Should it be the entire operating sys¬ tem configuration, all documentation, compiler, execution job language programs and data files? Or should it be just the changing elements of the processing? Once he decides what is to be retained, he must also decide how to retain it. A good outline of advanced techniques is avail¬ able in “Reliability of Real Time Systems” [60-65], 6.5. Programming Controls In line with the recognized objective of gen¬ erating technically sound programs, the ADP security program should include controls in the areas of program design, acceptance testing and standards. Each of these topics is discussed in the following sections. 6.5.1. Program Design There are five major program areas in which design can contribute to security. First is the inclusion of audit trails in the programming process. The basic objective is to make it pos¬ sible at any point in time to determine the status of a given piece of data. In most cases the systems analysts and system designers will want to involve the auditor in the design phase as he will be able to postulate the optimum placement of audit trails and controls. The second is the development of a test plan that will consider all possible elements of input, and the interfaces and operational aspects of each new program as part of the program design effort rather than as an afterthought. It is not enough to test a program for ranges of likely input; it should also be tested for im¬ probable, illegal and impossible input. In addi¬ tion, stand-alone tests usually are not sufficient to establish the adequacy of a given program or module. Not all programs need to meet the same test criteria; the stringency of the testing should be a function of importance, complexity and sensitivity. Development of written testing guidelines tailored to the needs of the ADP facility is an important step in achieving good control. The third control area is program change. Programs should be designed to simplify instal¬ lation of future changes. Every change, even those involving only one statement, should be authorized, approved, and documented with no exceptions. Otherwise, control is lost and the programming process becomes anarchistic. Pro¬ gram library maintenance packages, as men¬ tioned previously, can help in the control and maintenance of program changes. Naming con¬ ventions are essential to program change con¬ trol. The current trend is toward integrated data definitions for all ADP applications, so that every element will be unique. Controls on the accuracy of data records are the fourth design objective. There are a wide range of possible checks including keypunch verification, computer matching against prede¬ termined legal values for fields, self-checking digits and control fields. Standard design cri¬ teria should include the qualitative controls to be included in any new application or any re¬ vision of an old application. Finally, quantitative controls where feasible should also be installed during the design proc¬ ess. These could include control totals, run-to- run counts (hash totals), trailer records, dollar controls, automatic check-points/interruption routines, verfication of the output and input record counts and the like. Violation of qualita¬ tive and quantitative controls should cause error notifications maintained as an error sus¬ pense file. The need for quantitative and qualitative controls should be determined by the risk analysis. If the application is of high value, high risk, or consumes a great deal of ADP resources, these controls should receive more attention than low risk, low visibility applica¬ tions. 6.5.2. Program Installation One of the most sensitive points in the pro¬ gramming process is the release of an applica¬ tion to the production system, and its operation against a live data base. Installation of a new program should occur only after thorough pro¬ gram and system tests have been completed and approved. The more organizational entities participating in this approval, the better the control. The programmer, a testing or quality 60 [PDF page 63] PIPS PUB 31 Symbols: I: Initiates document A: Approves document U: Uses document - ■ i User i System Design Programming Job Control Operations ADP Management Feasibility Study Study Request I U A System Requirements I U Cost and Schedule A I A System Design System Flow Chart & Narrative A I U U A Source Documents & Input A I U U Record and File Organization I U U Output & Reports Formats Instr. A I u U U User, Keypunch & Control Instr. A I u U U Programming Program Narrative & Flow Chart I A Coding I Program Test Rec’s & Schedule I U A System Test Test Requirements & Schedule A I u U A A Test Report A A A I A A Implementation Implementation Schedule A I U A Data Conversion Plan I U A User Acceptance A I A Operation Program Maintenance I U U A Program Modification I U U u A Figure 18. Matrix of Suggested documentation to control and record programs. 61 [PDF page 64] FIPS PUB 31 control function, operations, and users should ail participate in getting the program from design to final acceptance test and into the live system. However, care should be taken to see that approval does not become a mere ritual. Each program should receive detailed, inde¬ pendent review. Larger ADP facilities may want to consider establishing a separate pro¬ gram test and control group. Smaller ADP facilities would probably be served adequately by defining specific procedures for the installa¬ tion process to be carried out by an existing group but with as much review and separation of responsibilities as is possible. Again, no pro¬ gram should be accepted without adequate and complete documentation which has been re¬ viewed and approved by an independent body. In case of disaster or non-availability of key programmers, the ADP facility could find itself quite vulnerable to loss if the documentation is inadequate. Figure 18 shows a suggested set of documentation which will provide the needed controls as well as technical information. 6.5.3. Documentation of Controls The procedural controls over data, operations, system design, programming and acceptance testing already described must themselves be documented if they are to be fully effective. This is often done by preparing documents called procedures manuals, operations and user handbooks, or similar titles. Responsibility for producing the documents may be assigned to a procedures group in a large ADP facility. The small ADP facility may call on individuals to document their particular areas. In either case, the ADP security planner should participate. He should analyze the security objectives of the ADP facility as discussed above to determine the role of the practices or standards in accom¬ plishment of security goals. Based both on these security objectives as well as on ADP manage¬ ment goals, a procedures program should be formulated for the ADP facility. An example of a table of contents for a programming proce¬ dures manual is included as Appendix C. 7. Security of Off-Site ADP Facilities 7.0. Introduction There are four basic reasons for making use of an off-site ADP facility: 1. The ADP needs of an agency are too small to justify an in-house ADP facility. A business whose routine data processing is done most economically at a service bureau serves as an example. 2. The efficiency and economy of the on-site ADP facility is enhanced by doing peak-load processing at an off-site facility. 3. A special service may be available from an off-site ADP facility which cannot be provided economically by the on-site facility. Use of an interactive time-shared computer for special jobs is characteristic of this usage. 4. In the event of catastrophe or major dam¬ age to the on-site ADP facility, critical ADP tasks are moved to a preselected off-site facility for back-up operation. The first three represent routine on-going use which is likely to increase over the years ahead. The fourth use results from the working of a contingency plan for an in-house ADP facility or as back-up for an off-site ADP fa¬ cility. What is recommended here is that the basic security considerations presented in these Guidelines for on-site ADP facilities be applied equally to off-site ADP. This chapter will ad¬ dress the problems that the ADP security plan¬ ner must face in evaluating the security of off-site ADP. Fundamentally, the user of off¬ site ADP is in a position very similar to the depositor in a bank—that is, the protection of one’s assets is turned over to another organiza¬ tion. Unfortunately, the user of off-site ADP does not have the protection provided to the bank depositor: the law, independent audit, and the FDIC. In fact, most ADP service bureaus provide a uniform (and often undefined) level of security at best for all of their users regard¬ less of individual user security requirements. As a rule the typical ADP service bureau does not guarantee any specific level of security pro¬ tection for users and does not accept responsi¬ bility for the losses that the users might incur because of data theft, processing delays or other disruptions. For these reasons it is not safe for the user to assume that work processed at an off-site ADP facility is being protected by adequate security measures. The conclusion is this: the fact that an agency does some or all of its data processing at an off-site ADP fa¬ cility (the operation of which the agency can¬ not control) does not relieve the using agency of responsibility for protecting its own data against loss or misuse and for avoiding delays in processing which interefere with accomplish¬ ing its mission. Indeed, the fact that the using agency cannot control security directly makes the analysis of security even more important. Therefore, it is recommended that an agency which uses off-site ADP facilities, support an ADP security program as described in this chapter. 62 [PDF page 65] FIPS PUB 31 If a combination of on-site and off-site ADP is used, then the person responsible for on-site ADP security planning probably should be re¬ sponsible for off-site ADP planning as well. If there is no on-site ADP facility, then the ADP security planner might best be chosen from the office responsible for vital records management, or the major ADP user in the agency. The designated ADP security planner should seek support and participation from all ADP users in the agency and advice and counsel from spe¬ cialists as suggested in section 1.3.2. 7.1. Analysis of Security Requirements While the basic techniques for risk analysis described in section 1.3 apply, the following approach may be helpful when off-site ADP facilities are being used: • Develop a loss potential estimate for the using agency as described in section 1.3. • Perform a threat analysis as described in section 1.3.2 but note that instead of a single environment (the on-site ADP facility implicit in the discussion in section 1.3.2), one must, in general, consider four different security situ¬ ations and environments as follows: 1. Protection of source documents, data files, ADP documentation data entry and out¬ put hardware, and related items while they are in the custody of the using agency. 2. Protection of data while in transit in either direction between the using agency and the off-site ADP facility. Note that data may be transmitted either electron¬ ically or physically (as source documents, machine readable media or output re¬ ports) . 3. Security of using agency ADP operations at the off-site ADP facility. The using agency may participate in an existing security program managed by the off-site ADP facility or may prefer to develop and maintain its own contingency plan to pro¬ tect its off-site ADP operations. 4. Protection of data, preprinted forms and other materials stored at an off-site loca¬ tion in support of the back-up operations plan of the using agency. • Develop an annual loss expectancy estimate as described in section 1.3.3. The basis for the estimate will differ from the single site situation in a number of ways. The using agency does not suffer a loss from the destruction of physi¬ cal assets (other than its own tapes, disk packs, etc.) at the off-site ADP facility. Similarly de¬ struction of data files and other material at the back-up site results only in the cost to replace them. These considerations are summarized in the table below for each of the five loss-poten¬ tial types listed in section 1.3.1. A Yes entry implies a loss potential similar to a full on-site ADP facility, a No entry means that the loss mechanism does not exist and the entry Minor refers to a loss limited to the relatively minor cost to replace data, documentation and related items. The ADP security planner should test the validity of the assumptions in the table for his particular situation so as to be sure that his loss expectancy estimates will include all sig¬ nificant factors. Potential Loss Type Loss Location On-Site In Transit Off-Site Bock-Up Site e Physical Loss Yes0 Minor Minor Minor Data Loss Yes Yres Yes No Data Theft Yes Yes Yes Yes Indirect Theft Yes Nob Yes No" Processing Delay Yes lTes Yes No* n The potential is probably much lower than for a full on-site ADP facility since hardware is limited to remote terminals. b It is assumed that tampering with data in transit would not go undetected. >' It is conceivable that an embezzler might be able to tamper witli inadequately protected back-up files and then destroy on-site files to force the use of the back-up files. However, this seems to be a rather farfetched fraud scenario. 11 If back-up materials were destroyed by the same event as the operational site, i.e., at the same time, a processing delay would occur. Hopefully, the back-up site has been selected to minimize the probability of a joint disaster as might occur if the operational and back-up sites were located on the same earthquake fault line. '■ Note that the using agency may elect to use its own facility to store materials to hack up operation at the off-site ADP facility. 7.2. On-Site Security Analysis of the security of the on-site por¬ tions of ADP operations is conducted as has been described in the preceding chapters of this handbook. Obviously if processing is done off¬ site, the ADP security planner need not concern himself about protecting an expensive, complex ADP facility, but he will want to consider points like these: • Physical protection, access controls and data controls for source data at the point where they are concentrated enough to become a tar¬ get for wrongdoers or where responsibility for data integrity shifts from users to ADP op¬ erations. • Protection of remote terminals against threats such as misuse or sabotage (deterred by physical access controls), damage caused by fire, flood, etc., or delays in processing caused either by physical damage to the terminal or by interruptions to electric power or communica¬ tions circuits. 63 [PDF page 66] FIPS PUB 31 • Physical protection for data files, documen¬ tation and other back-up materials which may be stored on site. 7.3. In-Transit Security The security analysis should consider the ex¬ posure while data and documents are in transit. Except for interception of electronic data trans¬ mission which is excluded from the scope of this handbook, the following’ points should be considered: • Physical loss of input. Where the cost to reconstruct or the loss from delayed processing is significant, steps should be taken to permit prompt replacement of input which is destroyed or lost in transit. Accidental erasure of mag¬ netic media is unlikely and is easily protected against by using magnetically shielded ship¬ ping containers. Heat, x-rays, and radar are all overrated threats which can be managed with common sense precautions based on a technical report by 3M Company [19] and an NBS report [ 12]. However, there is always some exposure to these threats and to the pos¬ sibility that a shipment will be misdirected or otherwise go astray. • Physical loss of output. Output which will be in the form of printed or microfilmed ma¬ terial is subject to the same exposures as ordi¬ nary mail but it obviously can be protected by the simple expedient of retaining the output data file at the off-site ADP facility until de¬ livery has been confirmed. Alternatively, one might prefer to trigger replacement on a report of non-delivery. In other words, unless non¬ delivery (the less common event) is reported by a specified time, the off-site ADP facility assumes delivery has occurred and need not retain the output file any longer (although exception reporting in this case carries greater risk than reporting each delivery). • Protection against disclosure. The loss po¬ tential analysis may show that either input or output are sensitive and must be protected against wrongful disclosure. Presumably the de¬ gree of protection required can be related to the value of disclosure to potential perpetrators and to the level of effort they are likely to use. Protection techniques used for classied ma¬ terials while in transit can be used as guidelines for developing protection techniques for un¬ classified but sensitive information. • Protection against tampering. The loss po¬ tential analysis may show that either input or output is subject to tampering for fraudulent purposes. Protection of input can make use of the same controls, in general, as are applied to in-house processing. However, one must take pains to see that steps are taken to protect not only input data, but control informaion as well. This is because one might conceal input data tampering by compensating changes to control data. Ideally, control information is kept on-site and output is not released until it has been verified against the on-site control data. However, if time constraints require verifica¬ tion at the off-site ADP facility, then control information can be protected while in transit. One may depend on the deterrent value of ulti¬ mate, if delayed, detection of tampering through later on-site confirmation. The ADP security planner should bear in mind that in many instances frauds have been concealed by substitution of altered output. For example, a recent report described how diver¬ sion of funds from dormant bank accounts was concealed by sending altered statements to the dormant account holders. The fraud was dis¬ covered when a delay in processing prevented the embezzlers from making the substitution.* This episode points up the situation where the fraud is revealed only by detailed output re¬ ports and so may be concealed (for a time at least) by tampering with these output reports. It seems likely that output which is shipped from one site to another for distribution, would be particularly subject to substitutive tamper¬ ing. 7.4. Off-Site Security The same technique is used to analyze se¬ curity at an off-site ADP facility as has been described for an in-house ADP facility but with a variation in emphasis as a result of variations in the loss potential. For example, if we estimate that we will operate 0.5% of the time at the back-up site, delayed processing losses would likely be on the order of 0.5% of their equivalent at the ADP facility normally used. In other words, the less likely we are to be operating at the back-up site, the less sig¬ nificant its reliability is to us so that we can place more emphasis on such factors as avail¬ ability, process integrity, technical compatibility and convenience in evaluating it. The emphasis for an off-site ADP facility which is used regularly would be the same as for an on-site facility, with the exception that one’s concern is obviously limited to one’s own assets. The ADP security planner can begin his security analysis of the off-site ADP facility by reviewing as much of the following documen¬ tation as is available from the off-site ADP facility: • a copy of the latest risk analysis. * "DP Figures in Bank Loss of $128,000,” Computerworld, I>. 1, February 3, 1073. 64 [PDF page 67] FIPS PUB 31 • a copy of the contingency plan—when it was last updated and the last time it was tested. • a copy of the last security audit, its date, and who performed it. • a copy of the security policy and proce¬ dures. • a copy of all other ADP physical security documentation. On the basis of the available documentation, an inspection and survey of the off-site ADP facility, and his own estimate of his agency’s loss potential, the ADP security planner should be able to draw one of the following conclusions about the off-site facility: 1. The security program at the off-site ADP facility is acceptable and no separate back-up arrangements are required. Presumably, the using agency will participate in and cooperate with the security program at the off-site ADP facility. 2. Protection of using-agency data and other materials is adequate, but reliability and con¬ tingency planning are inadequate; i.e. the ex¬ posure to processing delays is judged to be unacceptable. If the using agency finds that it can develop and maintain its own back-up plan, then use of the off-site facility could be justified despite the less-than-complete security pro¬ gram. However, the cost of the independent back-up plan should be factored into the price/ performance evaluation of the off-site ADP facility. 3. Security at the off-site ADP facility is judged to be inadequate. In this case it may be possible to arrange with the management of the off-site ADP facility for either a general up¬ grading of security, if that is what is needed, or installation of special measures for the using agency, such as special handling of using- agency data. However, when management is unwilling or unable to upgrade security, the using agency will have to look elsewhere for ADP services. When the risk analysis has been completed and an off-site ADP facility is selected for use, the using agency must support its ADP se¬ curity program as described elsewhere in this handbook. Specially, the following should be covered: 1. Security policy and procedures should be documented. 2. Using agency personnel who have ADP security responsibilities should receive appro¬ priate indoctrination, training and supervision. 3. An ADP security audit program should be established. The using agency may find that it can place reliance on audits performed by the off-site ADP facility for part, if not all, of its audit needs. 8. Contingency Planning 8.0. Introduction Each agency of the Federal government has an assigned mission. Plans are prepared and executed for the accomplishment of that mis¬ sion. These plans assume normal working con¬ ditions, availability of the agency’s resources and personnel and a tranquil community atmos¬ phere. Even so, the ADP security planner recognizes that despite careful use of preventive measures there is always some likelihood that events will occur which could prevent normal operations and interfere with accomplishing the agency’s mission. For this reason, he should include contingency plans in the ADP security program. Three different types of contingency plans are required for an ADP facility: Emergency response. There must be proce¬ dures for response to emergencies such as fire, flood, civil commotion, natural disasters, bomb threats, etc., in order to protect lives, limit the damage to property and minimize the impact on ADP operations. Back-up operation. Back-up operation plans are prepared in order to insure that essential tasks (as identified by the risk analysis) can be completed subsequent to disruption of the ADP and continuing until the facility is suffi¬ ciently restored. Recovery. Recovery plans are made to permit smooth, rapid restoration of the ADP facility following physical destruction or major dam¬ age. 8.1. Preparation of Contingency Plans Because good contingency planning is an im¬ portant contribution to stable ADP operations and will require substantial effort, it is recom¬ mended that a formal task force be estab¬ lished with well defined goals and a budget and schedule as a part of the security program implementation described in section 1.4. Fur¬ thermore, it will be necessary to have the participation of qualified people from other areas. Figure 19 suggests how tasks might be set up and assigned. Of course, each ADP facility will want to adapt to its own special circumstances and make full use of the resources available to it. 65 [PDF page 68] FIPS PUB 31 TASK FORCE MEMBERS Tasks ADP Management and Security Planner User Representatives c O' u G G £ C Building Security Agency Legal Counsel © Vj V 5