ursuantto E O 13526 MDR Case # 5477 rnUlUUmrnUll1 f5 $lBrnrnUU Ulf $rnlB Ul V ffi lIJUjffi ffi OO15GBPDlD15 f OOGBPDUj l GBPDUJlD 2nd Issue 1989 L 86-36 IN HONOR OF WOMEN IN SCIENCE AND ENGINEERING o o Elizabeth Rindskdpf o o 1 '1 ABOUT CLASSIFICATION o o o o o o o o o o o o COMPUSEC POLICY FOR THE INTELLIGENCE COMMUNITY ACROSS THE POND o o o o o o o o o o o o o o ' A NOTE TO CONTRIBUTORS o o o o o o o o o o ooooooo THE EXCITEMENT OF INFOSEC o o o o o o o o o HOW TO BUILD A USER-SEDUCTIVE ARCHITECTURE GOLDEN OLDIE o o o o o o o o o o BELATED THANKS o o o o o o ' CAMBODIA IN PEACE AND WAR o o BULLETIN BOARD o o o o o o HARDWARE REVIEW OCR DEVICES FROM THE PAST o o o GETTING STARTED o o o LETTERS o o o o o o ON THE LIGHTER SIDE ffl i o o 3 o oo oo o ' oo 4 '1 1 1 'fIllS BOetJMBN'f eON'fAiNS CGBB' ORB MAT RIAb SECREtNOT R L ASABL TO oo oo ooo o 7 10 11 i 13 o 16 17 19 28 29 37 38 39 41 CbASSI I D I BY SA t SSM 123 2 DECLASSIFY ON OFi iRatiRg A eRey's DeteFFFliRatieA Re uiFea CONTRACTORS DOCID 4010026 Published by Pl Techniques and Standards vot yr Noo 2 2nd Issue 1989 11 - PU BL1S H-e R o I I I I I ____ BOARD OF EDITORS Editor o o 1 Computer Syst il1sc ' Cryptanalysis o Cryptolinguistics ' --' ---' 963-1103 963-1103 963-5238 963-4740 Index 1 963-4814 Information Science ' c 1 963-3456 Information Security 1 972-2122 t963-30S7 Language I Mathematics o 1 963-5566 t963-6430 Puzzles 963-4958 Science and Technology Special Research o Vera R Filby 968-5043 Robert J Hanyok 963-4351 Traffic Analysis 1 J Illustrators 1 1 1963-6234 'IL ---''T _ _ 963-3738 1 963-6423 To submit articles or letters by mail send to Editor CRYPTOLOG Pl NORTH 2N018 If you used a word processor please include the mag card floppy or diskette along with your hard copy with a notation as to what equipment operating system and software you used via PLATFORM mail send to cryptlg@barlcOS bar-one-c-zero-five note no '0' via ALLIANCE send to PLBROWN note all caps attn CRYPTOLOG Always include your full name organization and secure phone also building and room numbers For Change of Address mail name and old and new organizations to Editor CRYPTOLOG Pl NORTH 2N018 Please do not phone The first issue of CRYPTOLOG came out fifteen years ago It was designed to be a monthly written by technicians for technicians informal newsy controversial lively and timely The budget crunch did away with the monthly But thanks to you the readerwriters the other attributes still apply To be successful CRYPTOLOG must reflect current operational topics in a way that interests you and others So wrote General Wolff the then DDO in a letter of introduction to that first issue August 1974 He went on to say I hope that you will want to read it and will help to write it We are repeating his words for the benefit of newcomers to CRYPTOLOG You the readers are also the writers The vitality of the publication is due to the animated exchanges of views among readers And now for two high-tech events to mark this anniversary We are state-of-the-art ourselves as befits the high-tech sponsoring organization CRYPrOLOG can now receive contributions via ALLIANCE as well as over PLATFORM Instructions appear elsewhere on this page Also beginning with this issue we are moving over to desk-top publishing on the Macintosh as conversion software permits Contents of CRYPTOLOG may not be reproduced or disseminated outside the National Security Agency without the permission of the Publisher Inquiries regarding reproduction and dissemination should be directed to the Editor CRYPTOLOG has moved to OPS-l 2N018 All opinions expressed in CRYPTOLOG are those of the authors They do not represent the official views of the National Security Agency Central Security Service FOR OFFICIAL New Mailing address Pl NORTH J 1i2 Ol fLY DOCID 4010026 liza th Ri7 4 skopf J ------------------- his article is based on an address at a luncheon the Federal Women's Program honoring NSA's l omen in science and engineering I ------------------- It is an honor to be asked to speak to you today but rather daunting too-I have read your bios and know the talent that I have before me Indeed such professional distinction and accomplishment - and in such great quantity-makes me wonder just what I might have to say today which could be directly relevant to you in your busy lives In short I asked myself as I sat down to prepare these brief remarks what message might I want to deliver this morning and what could I hope to accomplish in speaking to you The answer that came to me is simple and intensely personal It may be that what I most want to say will have greater meaning for me than for you And so I ask your indulgence and permission to express my own personal feelings on the meaning and importance of women in Science and Engineering today For what I want to say today is first thank you and second keep it up and third spread the word Let me spend a minute telling you a little more about each of these messages Thank you Some of you may recall an interesting program produced and directed by Peter Ustinov several years ago on the dangers and likely impact of a nuclear holocaust It combined multiple interviews with those involved day-to-day in manning - and I use that word advisedly - our nuclear defense Interspersed with these interviews was a dramatic portrayal of the impact of a nuclear attack on a small group of people-a mini day after if you will In all of the many interviews not one woman appeared never mind spoke The only woman in fact was one actress in the dramatic inserts I was struck by this total absence of women Weren't we half of the population Where were we when decisions about matters so central to the future of the nation- our civilization-were being made Shouldn't we be involved with our unique feminine perspective Of course we should Well why weren't we there Was Don Regan really right when he commented that women simply weren't interested in disarmament talks and related issues No I felt in my heart that that view was wrong The correct answer was a different one Women as a group have traditionally lacked the technical training to be able to participate effectively in many of the technical discussions surrounding issues so critical to the nation's security issues such as nuclear war disarmament and so on And lacking technically informed voices they have not been heard at all To my mind this is not only wrong it is wasteful and foolish as well Wasteful because we need all the good minds we have to solve the problems that confront us as a nation and as an agency Let's be frank This Agency and this nation are on the brink of a serious shortage of mathematicians and technically trained people To survive in the next century the nation must muster every bit of its technical and scientific strength And that means 2nd Issue 1989 CRYPTOLOG page 1 peR ePFICtAL USE m Ll' DOCID 4010026 P L 86-36 Their relative youth brings me to my second point Keep it up We need you now doing the creative work of the technical expert We will need you tomorrow as part of the Agency's Senior Management team And here 111 offer another personal perspective Too many times I find myself one of a handful of women if not the only woman sitting around the table in the Director's Conference Room I've formed a habit of taking a head count 75 men to 1 woman is not an acceptable ratio for these meetings It must and it will change but only if you make it happen NSA Women Honored for Contributions in the Fields of Science and Engineering Women's History Month March 1989 women must be a part of the solution without them there may not be a solution Foolish because I believe that women are in fact different than menneither better nor worse but with an important and often unique perspective and wisdom all their own We waste a precious asset when we organize ourselves to make important decisions without the presence feminine character and point of view And as a lawyer I might also add that we undercut the very genius of our democracy-a system based on the strength in diversity and the healthy tension in the competition of different ideas different strengths and different perspectives In short we need the participation of women in all avenues of public life But they cannot participate unless they are properly prepared to do so And so today I thank you for the hard work you have done to obtain the training that will allow you to excel in the technical leadership of this very high tech and very important agency And let me say a word about the group of women we are honoring today Twenty-seven candidates were selected for nomination by their organizations They were mathematicians computer programmers electronic engineers some had doctorates one taught mathematics at a school no less prestigious than the University of Texas Others taught themselves what they needed to know right out of high school Let me not be guilty of suggesting that you will have an easy time as you grow and progress in your chosen field I can guarantee that you won't As pioneers each of you will face special challenges many of you will lack adequate role models and as a result you11 be required to spend the extra energy demanded of all those who serve as our trailblazers I can tell you this with confidence because I've been there Just remember that your task is in many respects to change an entire culture that fails to encourage and may actively discourage the woman scientist or mathematician And many will have the added challenge of working two equally hard jobs raising young families right at the point when your career is at its most competitive For this you will require hard work military discipline and most importantly a sense of hUmor But keep at it you can do it Perhaps a phrase from my high school Latin course will help POSSUNT PROPTER COGEUNT - They can because they think they can Before I leave this point let me mention something you won't have to fight NSA Management I believe you have as supportive an environment as you are likely to find anywhere today Does that mean I guarantee you won't ever face discrimination No But I believe you will have the management support you need to Yet despite this highly impressive array of educaconquer the problems you confront Keep at it tional credentials and superb work performance all you won't regret it if you do but we will if you ar at least from my perspective -young between don't the ages of23 and 37 to be precise Lastly spread the word We need more technically versed women here at NSA and nationally 2nd Issue 1989 CRYPTOLOO page 2 FOR OFFIEURlA USE OPJL T DOCID 4010026 I hope you11 encourage other women in their pursuit of technical and scientific careers both here at NSA and outside as well And I hope you'll get involved when you can in serving as role models and teachers of girls and young women still in school And the earlier they see your example the better They need you to see what a kind of career the technically trained women can have Some will need your help to tackle the persistent problem of math anxiety and other difficulties particularly faced by girls and young women who pursue technical careers I believe much remains to be done in learning why it is that our young women find themselves considered as a group less comfortable than their brothers in learning math and science Do they learn the principles of math and the sciences differently than boys Possibly Are they often intimidated and discouraged when they confront technical subjects Probably Can you make a difference in attacking this problem I believe you can Your example inspires it encourages and creates the self-confidence essential to any fine performance Seeing your pictures as I came in today reminded me ofjust how powerful an example you can set Some months ago I visited the college my daughter has selected Some of you may know Wellesley and its long tradition of encouraging academic excellence in women For me it was a new experience but one I will long remember As I walked into the library's main reading room I looke d up to the clerestory windows bathed in lovely fall sunlight There hung all around the room were the elegant oil portraits of all the college's past presidents and each and everyone had the face of a woman I realized then with a thrill of surprise that in all my time in various fine educational institutions I had never before been a room where those so honored were exclusively women The message was clear in those confident gazes They could do it Alloll t ft aSSifioa t iOn ' P L 86-36 lp05ISS0 L - It is widely recognized that the volume of classified material NSA handles is enormous It is therefore incumbent upon us also to recognize the importance of proper classification practices and to ensure that our actions reflect the sound judgment required Individuals responsible for making classification decisions should remember that overclassification increases our vulnerability in two ways o It increases the volume of documents to be protected o It degrades our awareness of documents which truly merit TOP SECRET classification The classification decision process includes four steps o An initial dec sion to classify information based on whether it concerns military plans weapons activities or intelligence operations that are classifiable o A decision concerning whether unauthorized disclosures of the information can cause damage to the national security o A decision as to whether the information should be classified TOP SECRET - information the disclosure of which could cause exceptionally grave damage to the national security SECRET - information the disclosure of which could cause serious damage to the national security or CONFIDENTIAL - information the disclosure of which could cause damage to the national security o A provision for an indefinite period of classification unless an original classification authority establishes a date or event upon which automatic declassification should occur Note that classification decisions cannot be made merely to conceal violations of law inefficiency or administrative error prevent embarrassment to a person organization or agency or to restrain competition 2nd Issue 1989 CRYPTOLOG page 3 FOR OFFICIAl USE O lI Y DOCID 4010026 I C I T Networks covers computer security Emphasis is placed on the uniformity of protection mechanisms which becomes extremely important when we talk about interconnecting Automated Information Systems AISs Emphasis is also placed on the protection of intelligence processed in AISs and networks Thus this DCID applies not just to all classified information not just to all sensitive information but to all classified information that involves or is derived from intelligence sources or methods The first DCID treating computer security as a separate topic was published in 1972 The DCID in effect before 19 July 1988 was dated 4 January 1983 but was substantially the same as the DCID published in 1972 Obviously there have been a lot of changes in the computer security field since 1972 and the current version of the DCID reflects those changes COMPUTER SECURITY POLICY FOR THE INTELLIGENCE COMMUNITY GENERAL POLICY GUIDANCE U This article summarizes the current national levelpolicy on protecting intelligence processed in computer systems or networks It also discusses a problem that has plagued system security officers accreditors and users for years how to determine the mode of operation of a computer system in the NSNCSS environment P L 86-36 BACKGROUND U The Director of Central Intelligence DCI publishes security policy in the form of DCI Directives called DCIDs pronounced D-skids These cover all aspects of security including personnel security physical security operations security and computer security DCID 1 16 Security Policy for Uniform Protection ofIntelligence Processed in Automated Information Systems and U The purpose of the directive is to establish long-term year 2000 goals and near-term requirements year 1992 intended to improve the security of U S intelligence processed in AISs and networks The goal is for all Intelligence Community AISs to become trusted systems incorporating trusted products by the year 2000 In the period leading to the achievement of the foregoing goal interim measures to improve security will be taken including accrediting or reaccrediting systems implementing minimum near-term security requirements and improving access control and monitoring ACCREDITATION FOUO Each AIS and network covered by this policy must be accredited to operate in one of the four modes of operation dedicated system-high compartmented or multilevel Accreditation requirements vary with the mode of operation The minimum levels of trust as specified by the DoD Trusted Computer System Evaluation Criteria or Orange Book to be achieved as a goal by the year 2000 are Dedicated Mode - C1 System-High Mode - C2 Compartmented Mode B1 and Multilevel Mode - B2 Of coUrse a higher level of trust may be mandated for a particular 2nd Issue 1989 CRYPTOLOO page 4 SJOiCRET P DOCID 4010026 MODES OF OPERATION r n All Some Some Some Access All All Some Some Clearance All All Some Need-to know S s rn All AIS by its sponsor or responsible accrediting authority The responsible accrediting authority for NSAlCSS is the NSNCSS Senior Computer Security Coordinator T03 MODES OF OPERATION U In nCID 1 16 the fundamental basis for deciding which protective mechanisms are appropriate for a given situation is a concept called mode of operation There are four modes of operation used for AISs and networks processing intelligence dedicated system-high compartmented and multilevel There are exactly two sets of information needed to determine the mode of operation ofa system 1 the classifications and categories of information on the system and 2 the clearance level formal access approvals and need-to-know of all users of the system These two sets of facts are the sole determinant of the mode of operation of a system FO gO as Special Intelligence or SD TALENT KEYHOLE or BYEMAN information or any combination thereof Those programs that require as a condition of access the signing of a nondisclosure statement are considered to be SAPls Therefore in NSNCSS programs in the Very Restricted Knowledge system VRKs and the GAMMA subcompartment are considered to be SAPls and are given the status of SCI compartments and subcompartments in terms of the security levels required for their protection to For purposes oftms policy statement categories of intelligence information include Sensitive Compartmented Information SCI and Special Access Programs for Intelligence SAPls SCI includes either COMINT sometimes referred U For a system processing SCI clearance is based on criteria contained in DCID 1 14 Minimum Personnel Security Standards and Procedures Governing Eligibility for Access to Sensitive Compartmented Information In NSNCSS that means a user must have either a green gold or black badge U The attached chart shows how user attributes need-to know formal access approval and clearance level for the information on the system determine the mode of operation of the system These modes of operation are described in further detail below 2nd Issue 1989 CRYPTOLOG SECRET page 5 DOCID 4010026 SECRE'f DEDICATED MODE MULTILEVEL MODE U When all users are cleared for all data on the system and all users have formal access approvals for all data on the system and all users have the need-to-know all data on the system the system is operating in the dedicated mode of operation The system is not required to provide any technical security at all All the security for the system is based on traditional security disciplines such as physical security personnel security communications security and so forth U When at least one user is cleared for only some of the data on the system the system is operating in the multilevel mode of operation The system is required to provide a high level of technical security in addition to the technical security needed for the compartmented mode systems operating in the multilevel mode must provide a mandatory separation of users and data based on classification and clearance They must also provide an extremely high level of access control and accountability e g logins and passwords biometrics etc and a detailed and reliable level of audit trails Even with all this technical security some SYSTEM HIGH MODE CU When all users are cleared for all data on the system and all users have formal access approvals for all data on the system and at least one user has the need-to-know only some ofthe data on the system the system is operating in the system-high mode of operation The system is not required to provide much technical security only a reasonable amount of user separation Systems operating in the system-high mode are usually also trusted to provide some level of access control and accountability i e logins and passwords and perhaps some level of audit trails Almost all the security for the system is based on traditional security disciplines of the security for the system is bas trgtg -3 6 tional security disciplines ACCESS BY FOREIGN NATIONALS COMPARTMENTED MODE U When all users are cleared for all data on the system and at least one user has formal access approvals for only some of the data on the system the system is operating in the compartmented mode of operation The system is required to provide some technical security over that needed for the system-high mode Systems operating in the compartmented mode must provide a mandatory separation of users and data based on some sort of formal compartmentation and labeling They must also provide a high level of access control and accountability e g logins and passwords biometrics etc and a detailed and reliable level of audit trails Even with this improved technical security much of the security for the system is based on traditional security disciplines CONCLUSION CD If you have had experience working with previous policy statements on computer security you will note that we have come a long way in defining the parameters and technical measures for protecting intelligence information If this article has sparked your interest I recommend that you obtain a copy ofDeID 1 16 and its supplement Security Manual for Uniform Protection of Intelligence Processed in Automated Information Systems and Networks dated 19 July 1988 2nd Issue 1989 CRYPTOLOG page 6 SECRE'f DOCID 4010026 SEl6RElq ' ACROSS THE POND I 331 P L 2nd Issue 1989 o CRYPTOLOG o page 7 SE6REq ' HA n b VB QQUI IT QI-IANNE bS ONLY 86-36 i i i P L 86-36 EO 1 4 c EO 1 4 d DOCID 4010026 SE6RE'f' L 86-36 EO 1 4 c EO 1 4 d 2nd Issue 1989 CRYPTOLOG page SEGRE 8 IIANBLE VIA GOMHCf' GIIAPHH3LS OPtLY DOCID P L 86-36 EO 1 4 c EO 1 4 d 4010026 SEeRE distance Most shops close at 5 30 p m and most restaurants and theaters open at 7 00 p m Surrounding Cheltenham are the beautiful Cotswold hills which provide an immediate retreat from the bustling rather crowded business and residential areas L ----I FOUO In spite of some of those morale dampers the workers I met were generally cheerful and dedicated to using their highly developed abilities to the fullest extent Personally I was treated very well by my British supervisors and co-workers and made many friends I rarely met anyone who tooted his own horn and I discovered that a major attribute of English reserve is patience which does not allow hasty judgments but gives people a chance to prove themselves D Outside of work I met many really' friendly English people in Cheltenham a heterogeneous town about 3 miles in diameter located almost 100 miles west of London in the county of Gloucestershire in the South Midlands I spent considerable time socializing with friends from work church the Royal Scottish Dance Society and other organizations D I traveled extensively throughout the country on coaches and in my Mini in which I could drive on the narrowest of country roads and park in the smallest spaces in the cities Armed with my AA Automobile Association membership atlas and tour books arid usually in the company of British friends I thoroughly explored the Cotswold hills and Gloucestershire home of the famous Gloucester Cathedral approximately 10 miles southwest of Cheltenham visited London Bath Oxford and Cambridge and traveled to the Lake district in the northwest Yorkshire in the northeast where l stayed at Menwith Hill near Harrogate Devon including Torquay in the English Riviera and Cornwall in the southwest Canterbury Cathedral Windsor Castle and Hastings in the southeast and the Isle of Wight in the south D All over Great Britain including Scotland and Wales and also Southern Ireland and the Channel Islands I saw everything from palaces and manor houses to tiny village pubs rich in historical significance nestled in magnificantly scenic countryside a photographer's paradise Although a few places such as Stonehenge now have fencing to protect them against vandalism the countryside is kept very dean and neat by residents and travelers as well as by organizations like National Trust and English Heritage Also much to my surprise and delight travelers even encounter a fair amount of sunshine from time to time D A geographic and social feature of England D I rented a 7-room house in Cheltenham of which I had not been previously aware was -with all the modern conveniences one needs the contrast between the North and the South and which was located within walking distance The north of England is rugged and hilly with of work and the town center In town were an plenty of good low-priced housing but a abundance of lovely shops with the latest shortage of jobs mostly in factories which are fashions especially reasonably priced woolens closing down in large numbers By contrast parks and gardens one with a boating lake the flatter more urbanized south east has everschools churches doctors and dentists a choice increasing high-paying jobs largely in new of grocery 'stores and produce markets international computer firms but only very bookstores camera shops sport shops British high-priced houses and hardly anything like our Gas and the Midlands Electricity Board banks apartments for people who would be just and administrative offices appliance centers starting their careers or would be willing to theaters restaurants antique shops and move there from the north anything a person needs all within walking 2nd Issue 1989 CRYPTOLOG page 9 SECRE W rQb VIA COMUl'P GIhUHlELS OPlLY DOCID 4010026 SBCRE'f' U Concerning prevalent English views of America I noted that many people do indeed consider our country as being somewhat similar to theirs but just more technologically advanced and less polished and less rich in cultural heritage I noted also however that a surprising number even the large number who like the Americans whom they have met also strongly identify our country with the crime aggressiveness extreme wealth and wastefulness neurosis and superficiality that is seen in the very popular soaps like Dallas and Dynasty police films etc which they receive from us Although their news coverage is quite objective some commentaries and documentaries when seen in light of these onesided films are often misconstrued by the populace to the detriment of our country U Many people attribute several of England's recent social problems to U S influence Some also act cool at least initially towards Americans especially in the vicinity of Royal Air ForcelU S military bases like Fairford apparently because of their sometimes justified stereotype of Americans abroad Nevertheless I met even more people who were fascinated with everything American and wanted to visit America or had already visited America and had a marvelous time In my experience the majority expressed admiration for the openness and friendliness of the Americans they had met Others even changed their minds and decided to include a visit to America in future holiday plans U My tour in Cheltenham was on the whole very beneficial and enjoyable Although it seemed to take a long time to adjust to the new surroundings and cultivate friendships the results were well worth my efforts I learned a great deal from the English people not only in office matters but also in human relations and I discovered tremendous opportunities for travel I highly recommend this tour of duty 0 r------ _----------- A NOTE TO CONTRIBUTORS Though CRYPrOLOG is an informal publication it closely observes the etiquette of publishing One of the provisions concerns anonymity According to that rule the Editor must know the source of the information but may not reveal the identity without permission and without identification any submission is discarded whatever its apparent worthiness CRYPrOLOG has followed this rule many times over the years So dear author if you've sent a contribution without identifying yourself you won't see it in print Don't bother to scan issue after issue for your submission You must make yourself known to the Editor No exceptions Submissions received without identification are tossed out If you have already submitted something anonymously resubmit it with your name organization building and secure phone number-that is if you want to see it in print NEW MAILING ADDRESS And please include along with the hard copy a floppy with specifics about the computer used the operating system and the word processing software CRYPTOLOG P1 NORTH Unclassified L _ _ _ _ _ _ _ _ _ 2nd Issue 1989 CRYPTOLOG page 10 SB6RIB'J' ILUtDLB VIA 60MUl'J' 6IL NPlELS OPihY 4010026 THE EXCITEMENT OF INFOSEC '--Some time ago while I was having lunch with the Director of Security of one of our NATO allies and we were discussing the rash of books on intelligence agencies such as the CIA and Britain's MI-5 and MI-6 that were flooding bookstores he asked Why aren't there more best selling books on INFOSEC I replied It's because the best days we have in INFOSEC are when nothing exciting happens in the outside world When we are successful which we are most ofthe time the result is a non-event During the Walker spy trial Earl Clark an NSA INFOSEC expert said Give me access to your codes give me access to your ciphers and you won't have any secrets INFOSEC has all the secrets of US national security as well as the secrets of NATO and those of our allies around the wo ld to protect The responsibilities are awesome On a good day for INFOSEC the externals are placid but make no mistake the internals are boiling That's the excitement of INFOSEC The internal story is unknown and it must necessarily remain so to the outside world It is possible however to give some appreciation of the scope of the INFOSEC task with respect to the various elements each fascinating in its own right which collectively must be integrated into the total security pattern which constitutes INFOSEC Consider the challenge to the cryptomathematician Design a cryptoalgorithm to encrypt our mos t s ensitive secrets and having encrypted 1D2 P L 86-36 them we will give the resulting text to our most mathematically and technically sophisticated opponents and let them subject it to their most high-tech attacks It must protect the information for decades against such continuous attack That's not all It must do this under the assumption that the opponent has the algorithm but not the key To cryptoequipmerit engineers we say Embody the algorithm in an equipment that is fail-safe and to the evaluator we say Analyze the algorithm and the cryptoequipment that contains it and give it a seal of approval Impossible as it seems it is a task that must be coped with successfully if we are to have the ability to securely command and control our forces and to protect our strategic interests There are many situations particularly in tactical operations where valuable information can be derived not by breaking the encrypted transmissions but by analyzing the stereotypic formats the quantitative message data and other externals The task of protecting against such exploitation is the domain of transmission security This is an entirely different type of challenge the searching for seeming minutiae that could actually be a bonanza to hostile intelligence services One aspect ofthis or for some an INFOSEC category of its own is providing secure sequences for ECCM transmissions that are secure against enemy analytical reconstruction 2nd Issue 1989 CRYPTOLOG page 11 FQR QFFICIAL USE m LJPY DOCID 4010026 Hostile intelligence operations can concentrate on the attack mechanisms of their choosing The job of INFOSEC is to protect against practical attacks A technically pervasive phenomenon a known physical fact oflife is that electronic and electro-mechanical equipment when processing information necessarily create emanations which can be detected if not protected against TEMPEST is the field of INFOSEC devoted to the protection against unwanted unintentional compronusing emanations The technical challenge to determine how best to detect such emanations to identify those that may be compromising and then to devise corrective measures is complex However the real challenge is how to determine the cost-effective compromise At what point have we made such an attack unprofitable INFOSEC is always involved in optimization trade-off's but it is a twoparty game of exceedingly high stakes TEAPOT is a recently coined terms fori another aspect of the compromising emanations problem The difference between it and TEMPEST is that the emanations are hostily induced by bugs planted in the equipment In the TEA- POT category is the widely publicized GUNMAN operation of the recent past a rare case of our sharing the excitement with the outside world In the GUNMAN operation we removed tons of equipment from our Moscow embassy and replaced it with clean equipment in one rapid move before the Soviets could react Physical security in INFOSEC includes the protection of the cryptomaterials the codes ciphers cryptologics keys cryptoequipment When you consider the high value the Soviets place on the acquisition of our cryptomateriaI coupled with the vast amount of codes ciphers and keys in hard copy form around the world you can appreciate the enormous size of the this task If protection breaks down security breaks down That is why INFOSEC is a top priority target ofthe Soviet espionage apparatus Personnel security goes hand in glove with physical security since it is this route by which physical security is often attacked There are no stricter security constraints and checks on any personnel in the U S Government than on those working at the heart of INFOSEC 2nd Issue 1989 The rapid expansion of computers and the field of information processingl1tlsenormouslycomplicatedthequalitative and quantitative problem of protecting classified and sensitive information and at an exponentially increasing rate The previously discussed aspects of INFOSEC as complex and challenging as they may be have trend lines and data bases helpful in planning If Communications Security COMSEC is in a state of combustion COMPUSEC is in a state of explosion Harnessing an explosion can be almost too exciting Again it is a game a deadly two party game with extraordinarily high stakes It is vital to know the enemy capabilities if we are to be successful in countering them effectively That is the field of threat analysis Doctrine provides the procedural and regulatory sinew binding the INFOSEC capabilities into a coherent body The production of literally mountains of codes ciphers keys and other crypto-material with the utmost of security and accuracy is fundamentally important to an effective INFOSEC program Each of these areas of INFOSEC is a story in itself INFOSEC is not of course an end in itself It is only useful when applied in communications and electronics systems This opens another whole dimension to the scope of INFOSEC It is absolutely essential for both systems security and for the effectiveness and efficiency of the systems into which it is integrated that the INFOSEC professionals not only fully understand the technology of those systems but also the operations those systems are supporting Thus INFOSEC professionals are spread throughout the world in a wide variety ofroles Take one example Imagine the situation where a satellite launch is on hold for some unidentified technical problem and your equipment is the only Government Furnished Equipment in the whole system Now that's real excitement And now the final INFOSEC role systems security evaluation In accomplishing this task all the above discussed areas and their complex interactions must considered Coupled with this must be the consideration technology the varied environments and the wide range of applications and the ever-present hostile threat This must be integrated assessed and a determination made to give the seal of approval to a system OK to pass our nation's most vital secrets in this system The pressures on INFOSEC are great the task seemingly impossible and the external recognition and rewards necessarily almost nil But balancing all that out is the EXCITEMENT OF INFOSEC CRYPTOLOG page12 FQR QFfi lCll b l ISIi QN bJPY aCID 4010026 L --_IUJi5 By integrating off-the-shelf software some standard Agency hardware and programs already developed by the R Directorate we developed the basis for the OPELINT Architecture that can be used at any level within the ELINT community It has new ideas capabilities and opportunities in which analysts and programmers can be imaginative creative and explore new ideas and techniques We incorporated into the Architecture some computer models that give program managers decision makers and analysts the capability to develop differentcollectionandprocessing strategies along with high quality computer simulations that give visual results of their decisions The system is based on high quality graphics designed for easy use with concealed text to explain only those areas where users might need more detailed information U P L 86-36 U The process seemed simple at the outset All we had to do was gather information on all the collection and processing systems th t provide operational ELINT put it together and print out a nice document that would probably end up being cumbersome out of date before the end of the week and whose only useful function would be as a doorstop As we were collecting information two things quickly became evident 1 most information was hopelessly out of date some as much as 6 years and 2 the scope of the Architecture meant that we were inundated with information U We have been able to do this without a multimillion dollar budget and without having to purchase large mainframe computers or write massive programs or arrange large scale contracts P L 2nd Issue 1989 CRYPTOLOG page 13 CONFIBENTML 86-36 Eo 1 4 c P L 86-36 DOCID p' o'T je26 et ll i'm ll'fIPiL ------------------- '1 We plan to develop a system whereby the user can provide information directly to the data base and after review and formatting updating will be automatic This will give the users the opportunity to be involved in the development and maintenance of the overall system U Then W45 was faced with a conundrum what Architecture do you use to create an Architecture U We can anticipate shortfalls by comparing current requirements with new target trends We then help planners and programmers make smart decisions about eliminating the shortfall Their decisions then becQme a new plan that is factored back into the original equation either as a new capability or as a future plan -fEURt The primary purpose of an operational ELINT Architecture is to provide a baseline from which decision m8 kers can make those smart choices U Once we deterrmned what we would include in the Architecture we had to decide on what form we would use to present it This became a key issue because the archi tec ture Architecture will be used 1 The art of designing and erecting worldwide-at different buildings 2 A structure of structures levels ofcommand and by collectively 3 A style style and method people making decisions of deign and construction 4 Any on various types ofELINT design or orderly arrangement systems The mass of inperceived by man formation collected to date indicatedthatthe finished The American Heritage Dictionary of product would be several the English Language large volumes of printed material Since the ELINT community is a vast network of commands and organizations writing pnntlng and distributing hardcopy U A useful by-product of the baseline Architecdocumentation would be a major effort similar to publishing a full-length book Instead we detures has beenithe creation of an ELINT Glossary As we culled information for the baselines we pubcided to place the Architecture online and prolished ELINT terminology and acronyms in a vide access through various means printed hardcopy document We are now looking for a way versions of the documentation would be limited to make an interactive glossary for our online verto a master copy and several hardcopies for the sion ofthe Architecture archives I U Once the objective A rchitectures are developed we must update them whenever there is a change EO 1 4 c P L 86-36 2nd Issue 1989 CRYPTOLOG page 14 CQN RI N TMb EO 1 4 c P L 86-36 aCID EO 1 4 c P L 86-36 4010026 cm FfBSN'fIAL -teT To provide guidance in the areas ofcomputer modeling interactive glossary development and database establishmen t W45 purchased the services of a special engineering technical assistance contractor The contractor may be able to use or adapt some computer models already on hand or under development by ROB and R3 thus reducing costs saving time and assuring compatability with other programs The final U The process seemed simple enough to us and we planned also to develop a floppy-diskette version that for tactical units in the field U Our real problems began when it was pointed out that while our idea was good there were problems with printing in hard copy the diagrams and charts essential to the understanding of ELINT flow charts organizational diagrams mission tracks and even hardware configurations Once again the hunt began for off-the-shelf software Instead ofincorporating diagrams with the Architecture we are now developing a graphics-based program that allows the users to call up concealed or suppressed textual explanations of each graphic In addition we are developing a program that will allow the users to capture a series ofgraphics and assemble them in any order they wish for briefings and presentations U All of this software currently operates on an IBM or IBM-compatible AT with an enhanced graphics capability In the next phase when we have found an imaginative programmer to do it we will convert the Architecture to run on the Agency's high-performance workstation Our development of an OPELINT Architecture does not end with the establishment of a baseline We now must find ways to make it compatible with existing Architectures develop a program to add an interactive glossary develop computer models and look at ways to use the Architecture to help planners and decision-makers better understand ELINT area in which we want to concentrate is presentation Once the ELINT baseline is finished and modeling is under development we need to organize the information to prepare briefings write new plans and help document new programs For this purpose we are purchasing software that will provide high quality graphic simulations of ELINT systems and run on an Agency standard high-performance workstation The system will output the graphics as a composite video signal By creating and editing video tapes we hope to develop a library of simulations that can be combined with still graphics for high quality presentations U In summary the ArchitectUre is based on three ideas l the development ofan operational ELINT baseline and objective Architectures 2 the creation of computer models operating on a community-compatible data base and 3 the development ofcomputer simulations that can be combined with other visual material for demonstrating new concepts to decision makers and fiscal authorities iEURt Although far from finished the OPELINT Architecture has already begun to pay offfor W4 We have been able to use the information developed to create a comprehensive Operational ELINT Program Plan and we have opened some new avenues in the areas of graphic presentations for operational uses There are still some areas we need to explore and that offer opportunities and challenges for imaginative analysts programmers and planners U It is my hope this article not only aids others in their efforts to develop an Architecture but also elicits responses from those of you who are working on existing programs that may aid us in the execution of ours r--------------- I New Address for CRYPTOLOG I PI I NORTH L I 2nd Issue 1989 CRYPTOLOG page 15 CQ lF ITUb I I I I DOCID 4010026 t30LDEN OLDIE REFLECTIONS ON SELECTED NILOTIC LRN6URGES Nuer which is spoken along the upper White Nile Has more complicated morphophonemic alternations than there are theories about Mona Lisa's smile Rnd it may well be that they can be analyzed in terms of a pattern But if so the system is about as compleH as the control mechanism of a space uehicle designed to trauel to Venus by way of Saturn It is also possible that Shilluk is the language I haue been looking for which is neither a contour nor a discrete leuel nor a terraced leuel but a split leuel tone language But in any case I am uery glad that Shillulc is not my own language Because if it were I would be wandering around naked with a spear in one hand and a nail through my lower lip for decoration R compleH of cultural traits of Which my wife would take a dim uiew because she is not fond of lethal weapons and is sure that a nail through the lower lip would seriously interfere With eHpectoration Rnd although I haue neuer heard Dho-Luo I am eHtremely dubious about the feaSibility of deueloping competence in it with the lesson manual entitled Dho-Luo Without Tears Because I haue neuer heard any Nilotic language that isn't at least as hard as saying Peter Piper picked a peck of pickled peppers after seuen beers In fact my sentiments on another can be summed up by the simple obseruation that Dinka Is a stinka In short although these and other languages of the Sudan and Uganda and Kenya are indisputably neglected In the light of all the euidence I haue collected I am not just before proposing a gouernment-supported Language and Rrea Center specializing in Nilotic Because euen I haue a feeling possibly brought on by aduancing years that there is such as thing as being too ruddy eHotic I I P L march 11 1301 2ND ISSUE 1989 CRYPTOLOG PAGE 16 OPPIOW USE ONLY 86-36 DOCID 4010026 BELATED THANKS U 25 26 Oceangoing type U 37-44 Oceangoing type U 77 78 P L 86-36 Oceangoing type For most of us us Sigint-oriented types at NSA appreciation for the many-faceted art of cryptanalysis is something that just seems to go with the turf We have to give it lip-service at least since after all we do work here Well the other day while cleaning out my desk preparatory to getting ready to retire something happened to drastically change my perspective I came across an old copy of Cryptologic Spectrum Winter 1978 which featured a series of fascinating articles on Ultra and the Battle of the Atlantic More than fascinating to me because one of them called The German View by Prof Jurgen Rohwer made me realize that CIA might very well have been the instrument for saving my own life-and those of a number of other nervous GIs aboard our troop ship-way back in World War II days in February 1943 Suddenly in flashback I experienced a visceral appreciation of what the art of cryptanalysis could really do worked out on a scale of life and death down Reading Mr Rohwer's article I discovered something none of us knew at the time Bletchley Park's OIC Operational Intelligence Centre -which we of course had never heard ofl-had begun to make inroads on the German Naval Triton cipher after a bad six months during the second half of '42 when our people had a decrypting blackout and the Uboats were intercepting about one-third of the running convoys At last as he says Bletchley Part succeeded in solving the main problems of the Triton cipher circuit and from the 5th to 28th February the traffic was decrypted with seldom more than 24 hours time lag A comforting thought-just in time to protect our convoy These CIA successes meant that the British Submarine Tracking Room at the OIC could get warnings out to reroute the convoys and keep us away from the U-boat patrol lines If we had known this then a lot of us aboard the tiny USS Chateau Thierry barely 10 000 miserable tons in size would certainly have Looking back on the Battle of the Atlantic from slept better But there was still a problem with my own perspective aboard a tiny Liberty ship the wolfpacks according to Prof Rohwer at this time the German B-Dienst the I can imagine few situations better calculated counterpart of the Ole at Bletchley Park was to instill a feeling of Man Against the also having its most successful period it could Elements Add to that the gnawing fear of das decrypt many of the routing and re-routing Boot--the dread of a torpedo coming from signals and many of the U-boat situation somewhere out there We were heading northeast in early February on a convoy out of reports sent daily by the Admiralty or COMINCH Commander-in-Chief Tenth Fleet Boston both seasick and scared en route for He goes on to say that even if only 6 to 10% of Iceland We heard that a previous convoy had been hit-the one in which the Dorchester with the German decrypts were done in time to be of operational value they could have important the three heroic chaplains aboard had gone 2nd Issue 19B9 o CRYPTOLOG o page 17 FOR OFFICIAL USB ONLY DOCID 4010026 consequences He then cites February 1943 as a good example of this -the fateful time when the Battle of the Atlantic was beginnihg to come to a climax and when our own little troop ship was riding into it Prof Rohwer cites a certain convoy HX 228 heading northeast about which the B -Dienst had decrypted a position report Might this have been us It was the right time and direction At this point Admiral Donitz's people ordered the wolfpack Ritter to move north to find them or us Fortunately our heroes at Bletchley Park decrypt the order-God bless 'em-and the convoy gets an order to move north of the Ritter patrol line Another piece of luck for HX 228-the U-boat command about this same time intercepted and decrypted position reports on two other convoys ON 166 and ON 167--heading west away from our convoy Group Ritter was thereupon ordered to change course and set up a new patrol line along longitude 30 West to cover the expected passsing routes of those two other convoys-as I interpreted Prof Rohwer's account What followed must have been one of the most dramatic scenarios that fate could ever devise-more gripping than anything Hollywood could have dreamt up-move and countermove -as Bletchley Part and the U-boat command decrypted each other's messages the hunted and the hunters maneuvering for position in a life-and-death chase through leaden seas Finally it came to a grim conclusion Prof Rohwer reports In a fierce convoy battle which lasted for six days and covered 1100 miles the U-boats sank 15 ships and lost two of their number to the counterattacks of the escorts Many died many were spared Looking back I wish I knew for sure whether HX 228 was really my convoy I do know I want to look for more clues in Patrick Beesly's much-praised book Very Special Intelligence The Story ofthe Admiralty's Operational Intelligence Centre 1939-1945 published in 1977 Mr Beesly was deputy chief of the Submarine Tracking Room of the OIC at that time Anyway regardless of the exact role that CIA played in getting our own particular convoy to Reykjavik safe sound and hungry I'm now totally convinced that innumerable Allied lives were saved because of the hard work of the people at Bletchley Park OP-20-GCthe US counterpart of the OIC and in other places where the cryppy's trade was relentlessly pursued during those trying days So 46 years after the fact I send belated kudos to all of them from a grateful noncryppy survivor In normal times we can only dimly surmise how our lives are daily affected by unknown unseen people and events But in times of war how much more dramatic and powerful if the curtain is lifted and we catch a glimpse of it And in the case of my own particular convoy who knows but that some strange quirk of fate might have decreed that a decryption by our adversary the B-D ienst could become a strangely crucial factor in getting the wolfpack off our trail and on to someone else's If so perhaps I could even say somewhat selfishly Vielen Dank to you too our one-time enemy On the basis of my own experience relived retrospectively I can now agree with Prof Rohwer on the very deepest level when he says o without the work of many unknown experts at Bletchley Park the turning point of the Battle of the Atlantic would not have come as it did but months perhaps many months later Solution to NSA-Crostic #68 1st Issue 1989 Finally and perhaps most important translators must learn to unleash their minds Rote translation works for some texts all of the time and all texts some of the time but not for all texts all of the time It is at this juncture that creativity - - the choice of right phrase or word in English to match the thought and flavor of the original -- becomes crucial L -_ _ t'The Creative Translator Collected Articles on Translation 1973 2nd Issue 1989 o CRYPTOLOG o page 18 FOR OFFI6IAL USE ONLY P L 86-36 40100 61 4 P L c 86-36 SECRE't' SPOKE CAm ODIA 10 eACe AroD P ' L 8 6- 3 6 l1JA - _ 1 ' U RAMBO UIKA '1 S-CcO After spending over 12 years on the Vietnamese Communist VC target I had the opportunity in July 1968 to assume the duties of Chief of the Cambodian Branch in what was then B12 I took on this assignment because I saw it as both a career-enhancing job and an opportunity to work in a new environment But I can remember my fonner Division Chief asking shortly after I took on this assignment How are things in Sleepy Hollow 8 CeO I must say that in July 1968 the Cambodian problem could not be compared in any way with the Vietnamese problem With the war going on in Vietnam and the US involvement steadily increasing Cambodia also known as Khmer was not a target of importance 1 brought about a great change in Cambodia in the way the target was handled here at NSA and in the Intelligence Community as a whole The following is a description of this change andthe developments we had to contend with to support our customers on a real-time basis I ' SUMMAgy OF THE EFFORT 8 GGO In July 1968 the Cambodian Branch consisted of three team Imost of whom were civilians This included Traffic Analytic LinguisticlReporting and ADP Support Teams Most of our effort was concentrated on the analysis processing and reporting of Cambodian Anned Forces FANK communications S-COO Cambodia at that time enjoyed a low priority and in many cases hard copy reports satisfied the requirements These were annual disposition of forces order of battle reports and periodic reporting on political and military activities and changes in the military organization and equipment Electrical reports were issued by field elements mSM-626 Saigon South Vietnam 1 nd u S Technical Research Ships as well as by IH2 I I The only significant requirements related to the use of Cambodian territory by the VC 8-000 My main intent in writiIlli this articl forces either as safe havens or for the is to provide infonnation that might be of vl j lue infiltration of supplies equipment and personnel through Cambodia into South to production organizations which may1J ave to Vietnam r face a similar target change in the future I dedicate it to all the men and women who worked on the Cambodianpr 1 lem both at NSA and overseas andwiUingly put in many ho1 l rs of their own time to ensure that the job was accomplis ed in a quick and thorough manner I will not mention any of them by name imply to avoid leaving anyone out EO 1 4 c P L 86-36 2nd Issue 1989 o CRYPTOLOG o page 19 SECRE'f'SPOlfE P L 86-36 EO 1 4 c DOCID 4010026 ClU1'1' any evidence to support this theory and reported electrically on any indications of Vietnamese forces operating in Cambodian territory 8-000 Examples of the FANK message preambles and HF communications groups are shown in Figure 1 -f86r Also there was relatively high interest in Cambodian insurgent activities Three insurgent groups the Khmer Serie Khmer Leou and Khmer Rouge Cambodian Communists were active at the time primarily in the Cambodia-Thailand border areas We often reported on FANK operations and activities of insurgent elements None of these insurgent groups appeared to be of any great threat to the Cambodian government or its armed forces 8 GGO In my opinion the FANK communications network was a relatively easy target to collect process and analyze Most communications were in HF manual Morse 1 I I The field sites forwarded most traffic by courier except that which was encrypted or they felt was possibly important enough to meet product reporting requirements TRAFFIC ANALYSIS 8POIUi CRYPTANALYSIS AND LANGUAGE EXPLOITATION proximatelYninety-five percent of the correspondence passed over FANK communications was French plaintext messages A small amount of Cambodian language material was noted as well as some encrypted traffic We were able to exploit all traffic with except that which was badly garbled LS eeeiThe LinguisticlReporting Team consisted of about ten French linguists and a couple of Cambodian linguists They were able to scan and translate messages of significance and flag them for inclusion in a machine data base for subsequent reporting in hard-copy reports The messages meeting timely reporting requirements appeared in electrical summary reports The messages put into the machine data base were flagged by subject so that a pull could be made on them when a hard copy report on a specific subject was in preparation 8-000 The traffic analysts working on FANK e FANK messages were normally communications performed the normal tasks less than a hundred groups in length and expected of them by maintaining continuity on generally limited to one subject This made it all targets of interest preparing net diagrams fairly easy to categorize the messages and and message address group logs and providing identify the originators and addressees Often technical support to field elements via technical specific addresssees were spelled out in the messages and hard copy working aids and preambles and or the beginning of the message COMINT Technical Reports texts So just as with the traffic analysts' job the linguists reporters had a relatively easy The FANK did not use signal plans as but heavy work load sophisticated as those used by the YC Most HF communication groups were simplex stars THE KHMER COMMUNIST PROBLEM using day and night frequencies international procedure signals and French language operator 8-000 The development of the Khmer Communist KC problem was a very gradual chatter They often compromised address one that took place over about a two-year groups used in message preambles and period By early 1969 we began seeing occasionally in operators chatter and gave reference in FANK communications to increased advance warning of callsign frequency changes KC activities in scattered areas of Cambodia This coupled with the fact that we were This supported by collateral information eventually able to obtain through the Defense indicated that the KC were growing much Attaches Office DAO in Phnom Penh copies stronger and were probably being assisted by of the FANK overall signal plan made the the YC Extensive collection efforts were traffic analysts task relatively easy Our only undertaken to intercept KC communications problem was dealing with the large volume of we suspected would be HF manual Morse which material and HFILYHF radiotelephone EO 1 4 c P L 86-36 2nd Issue 1989 o CR YPTOLOG o page 20 8 ClU T 8POK I r DOCID 4010026 SECftE'f SPOtiE 2nd Issue 1989 o CRYPTOLOG o page 21 EO 1 4 c P L 86-36 S EURRF3'f SPOl DOCID EO 1 4 c P L 86-36 4010026 SRCKR' SPOKR 8-000 By early 1970 we had not been able to identify any KC communications We suspected that if the KC were being assisted by the VC the Vietnamese would provide them with necessary radio equipment and probably signal plan and cryptographic training and materials and so the KC communications characteristics and procedures would be similar to that used by the VC Eventually this proved to be the case -fS6t- In late 1969 while conducting general search efforts at Udorn Thailand USM-7 intercepted an HF manual Morse link initially identified as VC I placed on obtaining information on KC activities Meanwhile the intelligence requirements on FANK cottununications lessened as the US increased aid to Cambodia and a larger contingent of US personnel was assigned to Cambodia They were to help train FANK personnel anll develop a more effective intelligence gathering operation and expedite the flow of intelligence information to FANK units involved in day-to-day combat operations 8 GOO During the next two years we found ourselves having to deal with a number of problems besides the isolation and development of the KC communications network The traffic analytic work force had to be trained to do more detailed analysis on the more complex and sophisticated KC communications the language pool had to be changed from primarily French to a Cambodian cadre and those involved in preparing product reports other than translations had to shift gears to produce more timely reports and as with the traffic analysts learn to deal with a target that was not going to reveal as much in message externals and text as the FANK did During 1970-1972 one of our primary goals was resolving our language problem This as you can imagine was not an easy task - - - - - - - - - - - - - - - - - - since good Cambodian linguists couldn't be 8 000 Because of the signal plan used on hired off the street The Service Cryptologic this linkl Iwe tried Elements were our initial sources of linguists and both INSCOM and NSG were able to to have these communications worked in the Vietnamese Office Some even thought that the provide assistance over a period of time They adjusted their language training program to link was actually down-working of a VC communications group which had most of its train more Cambodian and fewer French linguists and the Language Intern Panel was stations located in Cambodian territory able eventually to provide us with one Further some believed that the VC were running the show in Cambodia and that this linguist communications group served their command 8 OGO By late 1972 our Cambodian and control apparatus in Cambodia Later we language pool had groW'nl Iwhich concluded that the VC group probably served included a numberofmilitary personnel We the VC cadre who were helping the KC were able to hire some of these military establish their infrastructure and were personnel after they had satisfied their military providing communications and technical commitment Fortunately this was back in the assistance Nevertheless this initially proved military person could clear the days when a to be a stumbling block in our attempts to building on a Friday and come back as a move on with the development of the KC next Monday civilian the communications network 8-eeO By this time interest il Cambodia had increased significantly agthe KC military and political arms starte9putting more and more pressure on the Cambodian Government and its armed forces Intelligence requirements changed drastically with high emphasis bei ll g 8-000 During the next couple of years the Cambodian Branch was able to meet all timesensitive reporting requirements and at the same time develop a mean and lean work force By early 1974 the branch consisted of three teams I raffic Analysis 2nd Issue 1989 o CRYPTOLOG o page 22 EO 1 4 c P L 86-36 S CR T V01 1i DOCID 4010026 SI36RE Language Exploitation and Reporting The three teams had to work closely together to ensure that all communications were collected processed analyzed and reported on as quickly as possible S 660 Meanwhile an extensive collection and analytic effort had also been established at USM-7 Udorn Thailand where the bulk of the KC communications were being collected Every effort was being made to give the field the capability to perform first instance reporting on the target However because of language difficulties only a limited reporting effort could be maintained Thus most of the reporting effort fell on the Cambodian Branch at NSA The efforts involved in this process are described in the following paragraphs SPOIfS cover names and designators in the message texts for originators and addressees of messages as well as for political and military organizations The analysts had to be very careful when using the internal address information to co-locate terminals because it was often difficult to determine if a message was being sent by one control station to two or more subordinates or if the message was originated by a senior authority and being relayed to a second- or third-level subordinate The KC did not use broadcast communications to send a message to two or more subordinates the same message was sent separately to each 8 000 Another interesting traffic analytic problem developed a few months after the first KC link was isolated I TRAFFIC ANALYSIS I Also different message preambles As I indicated earlier the traffic analysts -- -I containing what appeared to be routing indiworking on the Cambodian problem had to cators started to be used on messages passed significantly alter their method of analysis in over the link The traffic analysts as well as order to contend with the KC communications the linguists were miffed by this development The KC communications groups consisted of two and it wasn't until we obtained some HF or more complex links which changed callslgns direction finding DF results that we started to and frequencies on an ad hoc basis Fortunately get a hint at what we were dealing with not all links changed at the same time so it was possible to concentrate on certain areas following signal plan changes iSer The KC HF manual Morse network grew gradually as the KC established their infrastructure in different parts of the country This allowed us to develop the communications while teaching the analysts to perform more detailed analysis Since the KC used complex links the intercept operators often intercepted only one end of a link in many cases they didn't have the capability to copy both ends at once So it was a case of searching through the unidentified material we received in the Cambodian Branch as well as material held in the unidentified mes in the Vietnamese Office to find backlinkactivity A time-consuming task but often profitable Through up- and down-time matches and message servicing activity we were able to recover many backlinks and aid the intercept operators in collecting those stations passing message traffic SO Because all KC traffic was readily exploitable the traffic analysts thought that their task would be relatively simple as it had beeJii with FANK communications However this proved not to be the case The KC used EO 1 4 c P L 86-36 EO 1 4 c 2nd Issue 1989 o ' R' PTOLOG o p geL23 8 6 - 3 6 S CU FF SPOI E EO 1 4 c DOCIO 4010026 P L 8I'lGRE' ' 8POI 1'l tS-CCO By the end of the war the traffic analysts had gained a lot of experience by working on a more difficult problem and learned how to use a variety of techniques to assist in their analysis This included not only HFDF results but airborne DF ARDF as well ARDF missions flown by US Army platforms over South Vietnam were able to provide accurate fixes on many of the KC targets while they flew missions along the Cambodian border In addition they were able to gain some experience in working on KC VHF LVHF voice communications which were occasionally intercepted by airborne platforms Unfortunately these communications were very difficult to intercept because of the low power and the type of equipment used by the KC and never really provided any meaningful intelligence information 8 000 All of the neVi linguists had been trained in Cambodian ti ing materials such as newspapers and books in C ambodian script and were taught to use Cambo dia dictionaries written in that scrip for Rqmanized Cambodian was not used in non-SIGINT material Thus to lone of the first things they had to do was to learn the romanized equivalents This was timeconsuming the more so because before these new linguists could translate the message they converted the romanized text back to Cambodian script by hand Another key problem was garbles that often appeared in intercepted material Even if there were only a few in a particular message the new linguists were unable to render the text readable I 8 GGO Another thing the linguists had to deal with was the different vocabulary used by the KC Besides using words not normally seen in FANK communications they also used a large number of abbreviations cover terms and cover names for persons places and things The linguists had to become familiar with all these before they could produce meaningful translations 8 GGO Figure 2 shows an example of a KC message preamble and HF manual Morse communications group note the difference between the KC and the FANK practises I 86-36 LANGUAGE EXPLOITATION 'ffleT To further complicate the linguists' task there was the problem of part messages KC essages were Ion sometimes more than 1000 rou s Moreover unlike the FANK the KC might cover a number of subjects in i me m sage Thus the linguists would have to deti rmirte the right order of the message parts befoi they could translate the message in the right context The linguists first scanned the cleaned up message forany information that met timely reporting requirements In most cases this was tactical inf lrI ation which proy ded data on KC attack plans OJ unit locations Normally they prepared a sho'rt narrative SPOT Report or CRITIC providing the inforynation to the DAO in Phnom Penh and our other customers then followed it up as soon aSillOssible with a full text translation The Defense Attache Office in Phnom Penh had a standing requirement for full-text translation or all messages containing L - ---- - -- -----I ----11 As we became more familiar with chart usage and patterns noted in the chart internals it was possible to develo process which na 0 -n l-y-di a n-o t Th-e T1-p T b-u t i t---- hindered our l ii guage exploitation efforts 1 I the linguists had develo Ped romanized equivalents for the C bodiah script characters patterning them a r th e' romanized Cambodian we occasionally n6tedin FANK Cambodian laintext messa es EO 1 4 c P L 86-36 2nd Issue 1989 o CRYPTOLOG o page 24 S ECIH 't' SPOftE EO 1 4 c P L 86-36 DOCID EO 1 4 c P' L 86-36 4010026 SHGREq ' SPOn E any intelligence information of value to them The DAO then sanitized the information and forwarded it to the appropriate FANK authorities for the necessary military action s-eeo Besides having to contend with all of the above the Language Team also had to provide on-the-job training for Cambodian linguists assigned to USM-7 Through early 1975 several groups of Army linguists spent ----1 three to _ _ -0 EO 1 4 c P L 86-36 CRYPTOLOG page 25 SEOaSq ' SPORE 2nd Issue 1989 4010026 SJWR 8Cf SPOIC 8 I o 10 '_J - ' ' r t ' - 113 5868 - i ' - v 1 T 3 5868 r 13 5868 113 5868 113 Se68 o 113 SE63 T13 ' SS6S EO LA c P L 86-36 - I I t J_ '' four months each in the Cambodian Branch prior to going overseas Much time had to be spent with these linguists to get them to the _ point where they could adequately handle the KC traffic sent back to USM-7 Although they did their best in the _-_ 2nd Issue 1989 CRYPTOLOG page 26 SECaBT SPOKE EO 1 4 c P L 86-36 ero 4010026 SBORE SPOliE -u I l' - '-_ - I i ' ' - ESSAGS OF AFPf1EC I AT io ' '- IT IS ALL Tee SELDCi ThAT PERSCNI El P FORMING AN INDISPENSABLE SU6JSCT _--''' ' TASK REcE IVE THE RECCGNHICN RIGHTFLilLYCESERVED - -- r Y CULc LIKf TO TAKE THIS CP QR CNITY Te EXPRESS N PERSC tL APPRECIATIoN Fci ' - eREli ' T I ON R lIE_S_fcR THE - h r ' Rt t d fir '1M ChiTiC o fdT25' riCKE o T' ' _ gf g g ASSASSIl hTEu MINISTERS ON 9 JUN 'AS A t r T' C'C' ''7' _ _ _ I T 'IC _ _ - o - -- r UST At D A TING EA E __ y6uMA TA- E SFECH L PRI _Y t v6 RS HHAVE DA LY CI'l CENDS IN TrE FIELC AS ELl AS IN PH ioi CPENH - - - ' '- ' - IfsELY c THAN o g XGDS-2 r 0 - -- t 000 2 - - _ - ' _ - ' ' f ' - ' - - ' '- _ __' _ __ __ - -_ - _- - __ -__ ' ' - _' _ 0-_ O' c f c V' i o t u __ - n -- Fer '- -' ------ 1 ---- field they never reached a level of proficiency sufficient to handle first-instance reporting on the KC problem 2nd losue 1989 o CRYPTOLOG o page 27 SJtC2ET VOKB DOCID 4010026 SECRET SPOKE REPORTING countering KC activities Unfortunately there were often indications that when FANK forces were tipped off to an impending KC attack instead of reinforcing their positions or conducting a counter attack they would simply withdraw from the area to avoid a confrontation Because of this and the constant heavy workload we often felt frustrated with the job at hand CCQ In addition to the SPOT Reports CRITICS and translations published by the linguists it was also necessary to produce summary reports on KC activities These reports dealt with message content as well as developments noted in communications A separate team was established to report these activities in addition to the Traffic Analytic Team One of the major problems we had early -f667- However there were times when our efforts were duly rewarded One prime on was identifying the users of the KC radio example was an event that occurred in early network reporters familiar with FANK practises provided specific information which June 1974 shortly after a number of Cambodian made it relatively easy to identify radio station Government officials were assassinated by the users This was not the case with the KC KC A funeral ceremony was going to be held in Phnom Penh which the U S Ambassador S-CCO I can remember many instances the FANK General Staff the Cambodian where we had long tough discussions on Cabinet and the entire diplomatic corps were whether we had enough information to at least scheduled to attend A KC message was tentatively identify a user on the basis of the intercepted indicating that they were going to structure of a communications group the area conduct a rocket attack on Phnom Penh during it was located in and the contents of message the ceremony We issued a CRITIC immetext Because the KC seldom identified diately upon recognition of the KC intent As a themselves as the FANK did it was hard to result of this CRITIC the ceremony as get reporters to go along with less than A rescheduled and according to the words of the validity But after a while this became normal US Ambassador your timeliness may have practice and to my knowledge the averted a catastrophe The Ambassador's identifications we came up with were pretty message to DIRNSA on this subject is shown as much on the mark Figure 3 CCQ The summary reports prepared by the Reporting Team were done to satisfy customer requirements not requiring real-time reporting This included periodic disposition of forces reports special topic summaries and inputs for the Southeast Asia SIGINT Summary and the NSA SIGINT Summary Often we were able to use collateral information to supplement what we obtained from SIGINT CONCLUSION 8 660 I hope after reading this article you will have a better appreciation of what the Cambodian Branch had to contend with during the war in Cambodia I wish I could conclude with a statement that would give a happy ending to this story But unfortunately the Lon Nol Government of Cambodia fell to the KC on 15 April 1975 We had given it our best effort as I'm sure the US personnel in Cambodia had done 8 CCO We published thousands of pieces of product with information which should have been of value to the FANK authorities in 8 G60 Needless to say the job of working on the Cambodian problem during the war was very rewarding from many aspects most important the privilege to work with a fine group of people During the period covered by this article B Group had a few reorganizations which fortunately didn't affect the Cambodian Branch The division we were part of was moved between a number of offices and changed designators each time--B12 B65 and B33 I B U L L ET IN B O A R D UPDATE ON CAPULET FOBO Individuals who have CAPULET software may obtain u dated versions from L -_----IB824 HQ 1A205 963-5184 CAPULET is a suite of CA programs for the IBM XT or AT 2nd Issue 1989 o CRYPTOLOG o page 28 SBCRE SPOIlE P L 86-36 DOCID 4010026 SIWRE'i' HARDWARE REVIEW OCR DEVICES P L Reviewed by 1 -- - -- - - Presented here are some principles of OCR our test methodology the results of our tests on three devices and of an accuracy test for several languages that was run on the best device ON OPTICAL CHARACTER RECOGNITION U The technology which allows printed matter to be transformed into computer digitized format is called Optical Character Recognition OCR There are two kinds of OCR software The first is dot matrix recognition which works by analyzing the location of dark and light spots on an image and matching them as closely as possible to a stored set of dots A more sophisticated technology is called pattern matching sometimes termed Intelligent Character Recognition ICR and relies on analyzing an image for its component parts much as a human does when reading Simply put a scanning device looks at the pattern for each character of text determines what this pattern represents and assigns a value to that pattern generally using ASCn text codes For example when the scanner sees a vertical bar crossed at the top with a horizontal bar it interprets the pattern to be that for the upper case letter T and assigns the appropriate ASCn value to the image U The accuracy of such interpretations is further enhanced to a limited degree by 86-36 IIAUDhB VIA jR8311 lexicons and artificial intelligence rules which help determine if a particular sequence of characters is in fact a plausible combination Artificial intelligence however is much less reliable than the human eye in viewing text as meaningful words A computer as well as a human can tell just as confidently that in English the character sequence happ should end with the letter y But the human can go one step further and also accurately fill in the missing letters for the phrase Happ N Year while the computer unless programmed with context-specific information probably cannot To a computer words such as Now Net etc are just as likely as the word New U R831 purchased for testing three optical character readers of the many on the market Although there are a number of different types of readers available including hand-held models desktop PC-based models and large stand-alone models the selection criteria were geared primarily to the need to digitize multilingual publications especially those having Cyrillic text Given these considerations the models selected were Intelligent Optics Corporation's lOC SPEEDREADER Kurzweil Computer Products' DISCOVER 7320 and Kurzweil's 4000 Intelligent Scanning System ISS DESCRIPTIONS OF THE DEVICES JOe SPEEDREADER U The IOC SPEEDREADER is a desktop pattern-matching page reader which can input text or graphics from a printed or typed page to a personal computer or word processor Any 2nd Issue 1989 o CRYPTOLOG o page 29 P L 86-36 SECftE'f GIIAUNElLS ONLY GO H i'f 'DOCID 4010026 SIWRR't' alphabet of a typewritten or typeset format can be recognized in a variety of fonts although point size is limited to the range 8 to 12 Scanning speed for high-quality text is approximately 30 characters per second for typeset and 40 characters per second for typewritten material In the graphics scanning mode the SPEEDREADER scans either line art or continuous tone images It has an automatic sheet-fed scanner which can be loaded with up to thirty pages of material Files of longer than thirty pages can be created using an append facility Minimum page size is 5 5 by 5 5 inches and maximum is 8 5 by 14 inches 1 U Character recognition is achieved when the scanner compares the patterns appearing in the text to a font selected from the document processing software and then provides the appropriate ASCII text code This font of patterns can be created by the user or chosen from a library of starter fonts contained in the software Creating a user font called training is a process which requires assigning an ASCII text code to each of the patterns appearing in the text This procedure might only require 30 minutes or so for very simple text but can be considerably longer for text having multiple typefaces The assigning of text codes to patterns is left to the discretion of the operator but is limited to two codes per pattern with a storage capacity of 256 patterns The operator might find this to be insufficient if as in the case of a document with several typefaces and or alphabets it is necessary to maintain typeface integrity in the output U In training SPEEDREADER the user must take time to account for all the characters and symbols appearing in the text This is important because if the scanner detects a pattern on which it has not been trained the output text displays either an incorrect character or a special symbol indicating an unknown character Since the SPEEDREADER has no capability to add to a trained font while scanning is in progress in order to correct this problem the operator has to stop scanning add the new character to the trained font and rescan the text U SPEEDREADER's output is a text file including the control characters inserted by the word processing program These control codes include hard and soft returns to preserve paragraph integrity but allow wordwrap functions tabs indents centering underlining space compression subscripts and superscripts SPEEDREADER supports most of the more popular word processing software packages on the market U All things considered the SPEEDREADER although the slowest of the three scanners tested performs satisfactorily as long as the input text is of good quality and contrast a caveat one could apply to all scanners Despite the limited capacity for trained characters and the lack of a concurrent editing function the SPEEDREADER is a flexible device which does an adequate job under most circumstances DISCOVER U DISCOVER 7320 is a PC-based menudriven pattern-matching scanner It reads both normal and landscape-oriented text in a variety of fonts ranging in size from 8 to 24 points A maximum speed of 50 to 60 characters per second can be achieved depending on the complexity of the character fonts and the quality of the scanned document Software containing system lexicons of approximately 50 000 words is available for English Spanish Dutch French German Italian or Swedish In addition user-defined lexicons of up to 10 000 words each can be created to supplement the system lexicon U DISCOVER is also an image scanner which can scan drawings and photographs with 15 levels of contrast control There is a built-in sheet feed which will handle up to ten pages' of material DISCOVER can operate in a background mode' under MS DOS thereby allowing document scanning and access to other software packages and PC functions simultaneously U Other features include an assurance threshold which is a system confidence level telling the software how sure it must be of an identification before considering a text character as being recognized a window option which allows the operator to specify a zone whose contents should be captured when the page is scanned and a column recognition function which allows the system to analyze the pa es for the specified number of columns and separate them into discrete zones U DISCOVER 7320 is probably the fastest scanner and the easiest to configure but it is the least flexible It is not trainable and can be used realistically only for monolingual 2nd Issue 1989 o CRYPTOLOG o page 30 SFJ 6 RS'f' IIA mIsS qA 60MHt'f' GILUHtSIsS OPtLY DOCID 4010026 SEGRE' F documents containing text in one of the preprogrammed languages Also even though the DISCOVER 7320 reads different typefaces it gives no beginning- nor end-of-font markers in the output text to aid parsing Nevertheless if scanning requirements are such that these perceived drawbacks are not important then DISCOVER 7320 can provide a reliable product KURZWEIL 4000 ISS U Kurzweil 4000 is a stand-alone patternmatching scanner which can accommodate multiple fonts within a single document and read virtually any type font in sizes ranging from 6 to 24 points Standard equipment features a display terminal hard disk drive and a floppy disk drive used for file backup Optional equipment includes an electronic tablet for document mark-up and a document feeder capable of holding up to 25 pages Input can be either from bound books or individual pages measuring up to 11 by 14 inches Separated pages can be placed individually on a glass surface directly over the optical scanning head or fed into the scanner by means of the document feeder In the case of bound books the pages must be placed on the scanning window Depending on print quality and format of the original material scanning speed is between 20 and 50 characters per second U The standard language configuration is for English but additional optional language packages are available for German Italian French Dutch Danish and Spanish Language packages come with lexicons containing approximately 33 000 words Since it is a trainable device it is also possible to program the 4000 to read other languages as long as the text is composed of discrete characters U Output is a file which can be transmitted to magnetic tape or hard disk displayed on the terminal screen or transmitted to a peripheral device using appropriate communications software The 4000 will also receive documents from asynchronous ports and magnetic tape machines A more detailed examination of how the Kurzweil 4000 functions follows below need to be evaluated before deciding which scanner will produce the most accurate and reliable output In addition to considering 1 the features and capabilities of the scanning device and 2 the format and clarity of the printed text the user must give equal consideration to 3 the amount and method of output editing to be done as well as 4 the format and intended use of the digitized product CU It is also possible to have a single restriction as the deciding factor in the selection of a scanner If for instance there is a need to capture a publication having significant intrinsic or monetary value removing the pages of the document from the binding is not practical Digitizing in this case is then limited to a device capable of scanning bound book material Only after all of the variables are evaluated can a user make a reliable determination as to the best scanner for a job 6 660 Performance evaluation of the three scanners was conducted on a wide variety of documents such as office memorandums monolingual and multilingual publications printer text and magazine articles In addition to English languages included Amharic Arabic French German Italian Polish Portuguese Spanish and Russian The quality of the material ranged from very good for clear high contrast print of a newer publication to very poor for older worn documents whose print has faded through use and age One goal of the evaluation was to judge the relative capabilities of the three scanners but the primary emphasis was on determining which scanner produced bilingual digitized output in a form most suitable for incorporation into a relational database CONCLUSIONS CD Given the rigid constraints of producing bilingual digitized text for a relational database the flexibility of the Kurzweil 4000 appears to make it the most effective device for the job This should not be interpreted as a blanket endorsement of the Kurzweil 4000 over TESTING AND EVALUATION the other scanners but merely a best-bet choice for this particular scenario Less versatile but CU The process of converting printed material nevertheless still a viable alternative is the into computer format requires more than simply roc SPEEDREADER The remaining scanning choosing a scanning device and feeding a device the Kurzweil DISCOVER 7320 is least document into it At least four mutually suitable for the purposes described dependent phases of the digitizing operation 2nd Issue 1989 IIA rQb-g 1fT CRYPTOLOG I' CKI' T page 31 GO lnTT GRA T TKbS O Tb y DOCID 4010026 SIWREJ DISCOVER 7320 I I I 'I I I I I I I I I I I I 1'1 I Model Type Cost Speed Font Size points Graphics Capability Scan Bound Book Language Lexicons Trainable Concurrent Editing Column Recognition Multifont Recognition Parsable Output Scan Bilingual Text 1 I I I 1'1 I I KURZWEIL 4000 I I Desktop I $11 950 150-60 cps I 8-24 I y I N 150 000 WD I N I N I 'I I I I I I I I J I I I I J I I I Desktop I $5 495 I 30-40 cpsl 8-12 I y Y N 33 000 WD N N y y I N N y I I I I Stand-alone $29 107 20-50 cps 6-24 y y y y y I IOC SPEEDREADER y N y y y y I I I I I I I I I The cost for the Kurzweil 4000 includes the optional electronic tablet $3097 automatic document feeder $2655 and four language lexicons $1425 each Table 1 COMPARISON OF OCR DEVICES U ESTIMATES OF ACCURACY U Table 1 provides an estimation of the accuracy one could expect from each of the optical character readers Percentage of accurately-read characters is based on results of test material or represents computations derived from a comparison of characters within scripts It assumes that all text is monolingual and of high quality PRODUCTION OF A DIGITIZED DICTIONARY 8 660 After preliminary testing of the three scanners was completed R83 decided to begin production of a digitized version of a text for inclusion in a prototype of the database system Callaham'S Russian-English Chemical and Polytechnical Dictionary was chosen as the candidate for this effort because of its high priority status in the queue of working aids for incorporation into the Idatabase system and because of its high-contrast good quality text Being a bilingual dictionary it also gives a true representation of the challenges R831 faces in converting hardcopy publications to relational database form I I U The Callaham is 852 pages long with two columns of text per page Most columns contain 62 lines of text and an average of approximately 32 characters including spaces P L 86-36 2nd Issue 1989 IIA BbB per line This is roughly 3 400 000 characters of input which produces output of approximately 4 000 000 characters Included in the output figure are begin-end font indicators and multiple codes for certain individual letters The text is consists of standard Roman English characters approximately 58 percent English italics 7 percent and bold Cyrillic 35 percent Special characters include diacritical marks acute accent grave accent circumflex tilde hacek diaeresis degree symbol percent sign virgule I ellipses C em dash superscript subscript and Greek letters U The first step in the procedure for scanning a new document is to define the initial set up parameters This process involves stepping through a series of menus in order to set the framework for training The user is prompted to define a document name training set name training mode lexicon page incrementation option and use of super sub scripts and underlines Some textual ambiguity information is also required in order to help the scanner distinguish between 0 0 111 and 111 U After the initial parameters have been defined a training set for the document is formulated Training is the process by which the Kurzweil system learns to recognize the set of characters which constitute the document text The machine is configured to the manual CRYPTOLOG page 32 SIilGREq' 'lIlt GOl H 't'f' GIIA Hf8ES OUEI' DOCID 4010026 training mode in which the operator's role is to review text character'by'character or line-by-line and verify that the images displayed on the screen are clear and well-formed Training should not be done on partial or ill-formed characters Those characters which do not meet the confidence level of the system recognition logic are displayed in inverse video and are called interventions All interventions should be verified as correct by the operator or assigned the appropriate identification All images not just interventions should be reviewed by the operator to ensure the training set is not corrupted with incorrect values progress An existing training set can be supplemented during production through the use of a special function key as long as the storage capacity of 432 images is not exceeded Exceeding that limit results in a corrupted training set D Classification of fonts is also done during training Fonts are numbered 0 through 9 and are flagged on the screen and in the output file as n at the beginning of the font sequence and n n at the end The predominant font in the text which in the case of the Callaham Dictionary is standard English is assigned font zero Font zero is maintained only internally and appears in the output file without a font flag Superscript subscript and underlines are flagged s s i i and u u respectively D When the operator is satisfied that most of the text characters have been incorporated into the training set it can be copied to system memory for use in production Training sets are stored in memory and can be used repeatedly as long as the characteristics of the text remain constant D With the establishment of a satisfactory training set the operator can now begin the production process by stepping through the menus and setting the scanning parameters Part of this procedure requires the operator to define page and column limitations by using the electronic tablet This entails marking the upper left and lower right corners of each D There is also an automatic training mode which is used to analyze and display characters to the operator once a base level of recognition has been established that is after a training set has been created and production is in DISCOVER 7320 I I 1 ' t I I I I I I I I ' I Amharic Arabic Chinese Czech Danish Dutch English French German Greek Hebrew Hungarian Italian Japanese Korean Polish Portuguese Russian Cyrillic Spanish Swedish I I I I I I I I I I I I I I I I I I I Table 2 I I I KURZWEIL 4000 I I I I I 0 0 0 60 85 95 95 95 95 50 0 70 95 0 0 65 55 5 95 85 I I I 1'1 I IOC SPEEDREADER I I I I I I I I I It' I 3 97 Estlmates of Accuracy SEGRE'f I 3 95 95 95 95 95 95 95 95 95 95 3 11 95 95 90 95 95 95 97 97 qA GOM1U P GHANNELS I 75 85 5 5 97 97 97 97 97 97 97 97 97 97 5 18 97 2nd Issue 1989 o CRYPTOLOG o page 33 IIA mLE I m LY I I I I DOCID 4010026 Sf36RBr p column of each page of text When all the page column boundaries have been defined the material is ready to be scanned The operator loads the pages of text into the feeder or places individual pages or bound documents directly on the scanning window and sets the optical scanning head into motion For the Callaham the pages were removed from the binding and scanned individually In this case at least placing the individual pages directly on the scanning window seemed to produce slightly better results than using the document feeder U Depending on the degree of accuracy required the operator has a number of options for scrolling through the text as it appears on the screen The SCROLL UNLIM option allows uninterrupted scrolling of text in the editor This is the quickest way to scan a document but produces the least accurate output since corrections changes or additions to the training set or output text can e made only bby accessing the file via a word editor or word processor U A second option is the SCROLL GOOD mode which allows scrolling to proceed uninterrupted as long as the rate of interventions questionable images is low If the rate of interventions increases operator verification is required arrow scrolling options thereby editing the document on a line-by-line basis during scanning increases the accuracy rate to approximately 96-98 percent But even such a modest error rate results in approximately 80 000 to 160 000 incorrect characters in the output Three different methods were used to detect and correct these remaining errors U The most efficient method is a computer program which is written to reflect inconsistencies in the presentation of the various elements of information contained in the dictionary For instance the Callaham has an identifiable finite number of abbreviations always appearing within parentheses which provide contextual information It is a simple matter then to write a computer program to verify that all abbreviations within parentheses match a list of context abbreviations Another way we used a computer program in editing was to develop a partial spell checker to help verify the orthographic sequence of dictionary headwords Approximately five percent of the errors produced through machine scanning were identifiable programmatically U Another technique in editing errors is employing the Search and Replace functions of a word editor The nature of optical character recognition especially when applied to a large volume of text often produces errors which U A third mode is the SHOW right arrow appear in a consistent repeated fashion It is option which allows the operator to skip from those errors which fall into this category that intervention to intervention and make can best be edited using the Search and appropriate corrections or additions Replace functions For example in the Unfortunately using this method alone does not Callaham the scanner often was unable to allow for correction of those errors not distinguish between the character sequences appearing as interventions a common f fi and r' when appearing in italic script occurrence Other arrow keys allow the cursor As a result the part of speech prefix was to be moved one position at a time and help frequently rendered as prefx It is a simple expedite the editing process matter then to globally replace all occurrences of the incorrect sequence using the U The final option is the ACCEPT mode SearchlReplace function The operator which scrolls text one line at a time The however must take care to ensure that the ACCEPT mode especially when used in sequence being replaced is not a valid sequence combination with the SHOW right arrow elsewhere in the text If global replacement is option and arrow cursor keys produces the not possible then replacing the sequence in highest accuracy output but requires the most question must be treated on a case-by-case time since the operator is verifying both basis interventions and non interventions EO GGOl Another instance of how the POST-PRODUCTION EDITING Search Replace function can be used is when text although technically correct in the output U It is estimated that unedited optical scanning of Callaham's Dictionary would result needs to be modified to accommodate a database design For example since the in an accuracy rate per character of 85 to 90 Callahan has two narrow columns of text per percent Choosing the ACCEPT SHOW right page words are often hyphenated When this 2nd Issue 1989 CRYPTOLOG SIilGlHJ'-P page 34 HANDLE VIA OOMIN'f' OHANNELS ONLY DOCID 4010026 SEGRf 'F P L 86-36 Qutput is incorrrated into a database such as for these hyphens must be removed so that an interrogation of the data yields comprehensive retrieval information This problem is compounded however by the fact that many words need to maintain hyphenation Further complication is encountered because the operator often cannot tell the difference between a word which should have the hyphen removed and one which maintains hyphenation To illustrate the average OCR operator may not know if a chemical term such as hexamethyl-enetetramine contains a hyphen because it is a compound word or if it is because of columnar limitation in the source document D Many different types of errors are encountered during digitizing but generally they can be broken down into two broad categories those caused by scanner programming and those resulting from problems with the input text The predominant error by far is incorrect character recognition that is the optical scanner detects an image but assigns an incorrect value to that image This occurs most often when the scanner is forced to perform a task it was not designed to do The Kurzweil 4000 is intended to scan monolingual text in English French German Dutch Italian Spanish Swedish or Danish D By trying to scan bilingual text the machine is being forced to recognize characters in a language for which it has not been programmed Nevertheless scanning even bilingual text yields a reasonably high accuracy rate as long as the character sets for both languages are based on the Latin alphabet For example scanning an Italian-English dictionary is not a major problem for the Kurzweil 4000 because both Italian and English use Latin script Italian has five characters which are not used in English and English has five characters which are normally used only in borrowed words in Italian publications Scanning could be accomplished by determining the predominant language in the dictionary initiating the lexicon for that language and teaching the optical reader to recognize the extra letters from the secondary language scheme of Latin characters This can create recognition problems for the scanner however since a single image might appear in both character sets and therefore have two values An illustration of this is the image c which appears in both the English and Cyrillic alphabets The scanner needs to assign one value when this image appears in Cyrillic and a different value when it appears in the English character set D Incorrect character recognition occurs also when images are very similar to each other The problem appears to be compounded in bilingual text when multiple language fonts are used and one of the fonts represents a nonLatin alphabet The Cyrillic letters il H II have similar images and are often indistinguishable to the scanner The result is an error rate of approximately 10 to 15 percent which is correctable mainly through manual scanning of the output text But recognition of characters with very similar images in the preprogrammed lexicon results in highly accurate output at least 97 percent D Another fairly common error encountered in processing the Callaham is misplacement of the language font indicators The Kurzweil 4000 often had problems determining when one font ended and another began especially with a CyrilliclEnglish italics sequence This is a very difficult type of error to detect in editing because the flaw does not always happen in a discernibly consistent fashion D Other errors caused by the scanner itself include incomplete character values i e only two characters of a three-position code are returned in output omission of certain characters spurious insertion of characters and lines of text skipped rare D The digitized output of the Callaham also showed a significant amount of errors which were more a result of publishing errors as opposed to scanner interpretation of images These problems included skewed text misprints and misspellings in the publication as well as downright erroneous information D Although the acceptable error rate varies D Accuracy begins to suffer though when it from document to document the operator becomes necessary to scan bilingual text having usually will want to identify and correct as many of the errors as is practical As the type significantly different alphabets such as in a Russian-English dictionary The only method of of error becomes more obscure the likelihood that the error will be found decreases scanning text presented in non-Latin script Similarly the amount of time it takes to find such as Cyrillic is to utilize a transliteration 2nd Issue 1989 o CRYPTOLOG o page 35 SE6RJ3 J IIMmLE VIA GO HN'f' GIIA HiELS ONLY 4010026 SECRE't' the error increases In a large publication such as the Callaham Dictionary for which a high degree of accuracy was necessary processing time scan and edit averaged approximately one hour per page of input This produced an extremely low error rate conservatively estimated to be less than one error per 1000 characters which facilitated easy assimilation into a relational database ERROR COMPARISON U Table 3 compares the type and approximate percentage of errors encountered while scanning the Callaham Dictionary as well as the methods used to correct them METHOD Percentage IOn-line During of Error I Production Error Category I I 1 I 1'1 I I I I 1 Incorrect Character Standard English Incorrect Character English Italics Incorrect Character Bold Cyrillic Incorrect Font 10 Incorrect Font Placement Character omitted Miscellaneous Partial value Spurious Characters Text omitted Publishing errors Illegible text I I I I f I 1 1 I 11' I Iff Iff Iff Iff f PERCENT OF CORRECTION Computer Program Word Editor f f f t f f f f f I t f f f' 1 90 5 5 15 70 15 15 75 2 80 50 15 25 5 25 2 50 75 75 25 20 20 25 5 5 2 3 Table 3 Error Comparison CRYPTOLOG has moved - L--- ---- '-- Mailing address CRYPTOLOG ----- Pi NORTH 2nd Issue 1989 CRYPTOLOG o page 36 SSeRS'i' H tNDLB Tilt G01JPYHN'f' OIIAUUELB OHL'I f I fIt DOCID 4010026 From the Past HARVEST Machine Support in the Publication of Working Aids to Consumers u -Many consumers require listings and references as working aids which must be published and updated periodica lly I f a listing contains 9 000 records for example 300 additions and changes resulting from current analysis may have to be made to the oasic listing every month Ordinarily the original listing is published and distributed to consumers once a year When additions s nd changes are compiled monthly they are machine processed against the basic file through some file maintenance program Work listing are then sent to the sponsor for the next month's updating The sponsor usually rewrites the curre lt changes and additions in a fonnat suitable for presentation to the consumer and proceeds to type his own multilith masters The consumer then has the burden of making pen-and-ink changes to his copy of the basic document Under a new HARVEST process the changes and additions ' after they have been used to update the basic file are utilized to prepare a set of supplementary multilith masters This is done by extracting from the newly updated basic file all records which have been affected These extracted records are then arranged in' exactly the same format and sequences as they appear in the basic document These by-product masters are then sent to the sponsor along with his new listings of' the basic file He needs only to add his introduction and release them for pUblication All monthly change files are retained in the machine system so that they can be accumulated to produce a cumulative master file with only the latest changes included This in effect produces for the consumer one current supplement containing all changes and additions to date Obviously he can then destroy his preceding month's supplement For him to find the latest in ronnation he need only look first in his current supplement If he finds the record he knows he need not look any further he has the latest data If he does not find it he must look in his basic document knowing that the basic entry is still valid This tecr nique does away with pen-and-ink changes Multilith ma sters as a rllle result in a rather bulky an l awkward document Since pen-ana-ink changes have been eliminated a use of multilith masters can also be eliminated by replacing them with #16 paper As many as 108 records can be lista l on the #1 6 paper whereas the usus l number of records on multilith is a ma xi mum of 72 I24 photographica lly reduces the #16 paper image to a 1 etter sized sheet - an 8 x l inch page T' 1e resulting publication is much more compact 8 'ld sab1 e For example in one publication the multilith method produced 534 pages The SSJJ1e product using the #16 paper resulted in a document contair ing only 152 pages Incidently in the multilith method 124 estimated that 19 1 man-hours were required and the cost of printing and binding was $1 757 33 to produce 533 copies Under the latter method the man-hours required were 52 at Il cost of $652 72 to produce the same number of copies One hundred and thirty seven man-hours were saved and $1 104 60 were cut from the production cost Accorong to 124 the image erea and number of copies should determine wen #16 paper should be used When the number of copies is small and the format of the report image can be accommodated on standard multilith masters the masters should be used Some working aids have -elaborate introductory narratives In some cases these ha been put on magnetic tape as a separate file from which multilith masters or #16 paper listings are made thereby saving the sponsor from typing his introd J ction The technique described in this article could be applicable to any machine process which has a file mainte anceprogram Reprinted from bits bytes C4 MACHINE PROCESSSING INFORMATION BULLETIN Vol I No 3 July 1965 c45l x-mo P L Henry E Riley C412 2nd Issue 1989 CRYPTOLOG page 37 FOR O FFICM I ISH Qi J JPY X-4221 86-36 DOCID 4010026 1 Q tt t lll fj- J tO ltQ d 1 My object is to Make a report on something that has been accomplished o o Call attention to something that is wrong o Suggest a better method or idea o Share a personal experience o Enlist support o Explain a process o Ask a question o o Defend a pr incip1e o o o o Report news or announce a coming event Recognize an achievement Amuse and entertain React to something someone else has written None of the above but something else namely 2 My working title is o A better way to _ o Hurray for The fallacy of o It I s time to o The scandal of Are we paying too much for - - - - - - - - - - - - - - - - - - - - _ DAfter _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ What o o o o o o o A proposal for What happened at _ Why I agree disagree with A funny thing happened on the way to None of the above but who wrote 3 I will consider this in my own mind an open letter to and will address to primarily to her him them 4 OThis would be a particularly good time for such an article to appear because or o This subject is timeless 5 0 I am especially well qualified to write on this because or on the other hand o I want to speak up on this even though I am no expert on it because 6 Some tentative suggestions for a final title are 7 I think it would be most appropriate OCRYPTOLOG o The Cryptologic Quarterly o CLAr ion 2nd Issue 1989 for publication in 0 The NSA Newsletter 0 PQS Update 0 Vox Topics CRYPTOLOO page 38 FOR OFFICIt' b T I I Q JL Y _ DOCID 4010026 To the Editor D A version of my trip report on the American Mathematical AMS meeting in Providence last August appeared recently in CRYPTOLOG and to my surprise cause a controversy Buried on p 23 in a description of a talk critical of AMS by former presidential science advisor Edward David was the following section intended as a sarcastic criticism of the mathematical establishment that runs the AMS LETTERS Mathematicians must find better ways to attract young people to our field - NOT as currently take it or leave it The young have been leaving it in droves for adjacent fields Most students think we aim at winnowing out the weak and grinding down the ungifted They are right Most outstanding mathematicians view the world differently from David not a mathematician There is only a comparative handful of people in any generation capable of making important breakthroughs in theoretical pure mathematics They are of interest to us The rest are a waste of our time Academic education should be a pump not a filter U My own views and David's are clearly represented by the last sentence The we our and us are meant to be the sort of outstanding mathematicians that have power in AMS and that constituted David's audience the people who are driving young people from the field Surely it was never intended that us would be taken to meanNSA D Neither I nor anyone who knows me will take it that I think of myself as in the category of outstanding mathematicians I saw enough of them in graduate school to doubt I would want to be one of them if I could While my brother-in-law in academia would not agree I personally see little reason for society to tolerate people who undermine the general quality of university education because they care only about obtaining breakthroughs in theoretical pure mathematics While exposure to people doing research is part of training new mathematicians it is only part of the process D I hope that this straightens out any misunderstandings I do have to admit that when I attend a math conference I always feel a bit of conflict between my role as member of rather mediocre abilities as any star of the AMS will freely tell me of the mathematics society and my role as government bureaucrat seeing the considerable failings of the professional quality mathematicians from society's point of view But I have long since stopped cult worship of the great men of mathematics and taken a broader view that mathematicians like everybody else must justifY their salaries by arguments that involve more than their talent at glass bead games In this I believe I echo Admiral Studeman's speech to the AMS in Phoenix David Harris R512 U We regret to report the death of David Harris a valued contributor to the technical health ofNSA An appreciation will be published in a future issue 2nd Issue 1989 CRYPTOLOG page 39 FOR OFFlCtAL USE ONVY DOCID 4010026 than on pushing a paper through the publication pipeline Consequently we have the situation U The anonymous Cryptopoem published in where many analysts are not submitting their the 1st Issue 1989 was in fact writ nbYc J papers for publication but are just making ometiIllElPriorto197i copies for their colleagues Such documents _ _ _ _ _IA532 form hat is a growing underground technical P L 86-36 library To the Editor I To the Editor EFOUO I mentioned in 1st Issue 1989 page 13 there is a copy available in the T53 Technical Library FANX II 968-8611 Accession Number S-170 832 There are also a few spare copies in the R51 Mathematics Library FANX III 968-8580 U If there is sufficient demand for this document we will look into another production run In the interim we see no reason why this document could not be reproduced as needed Please address your inquiries to the undersigned U If we want our technical documents to be formally published and available to the general NSA technical population we are going to have to make the publication process less painful As long as we force analysts into a publish or produce dilemma we are going to have a shortage of publications 1 - P L 15 86-36 Dear Charlie U Now that you're in PI you've landed in clo- ver documentation-wise PI has a technical series and encourages its people to write for it There's a minimum offuss and moreover each _ _ _ _ _IR51 FANX III publication can have its own distribution U Other organizations have technical series that are similar '--To the EdItor P L lafti le U My initial reaction tol Where are Our Textbooks CRYPTOLOG 1st Issue 1989 was that Mr Gaddy was out of touch with the operational world because there are plenty of good technical documents being written at NSA I mentally started ticking off textbooks thought pieces working aids and other types of technical documents whose absence Mr Gaddy was bemoaning Then I realized most ofthese examples had never been formally published U Why aren't a number of our best technical documents being published Quite simply the process is so lengthy and frustrating that many good analysts will not put up with the hassle These people value their time and talents and prefer to expend their efforts on technical problems rather IChief PI 86-36 CRYPTOlOG welcomes your letters and comments Send them to The Editor PI NORTH NOTE TH ENE W M Rill NG ROO RES S 2nd Issue 1989 CRYPTOLOG page 40 FOR OFFICMb HSE Ol 'I JPY DOCID 4010026 On the Lighter Side Introducing Exciting Friedman BINGO Another boring briefing in Friedman Never again with the POLEMIC's new FRIEDMAN BINGO game guaranteed to keep even those with the heaviest eyelids awake and on the edge of their seats during any talk Just cover up a square any time you see one of the events on the FRIEDMAN BINGO game Get five in a row horizontally vertically or diagonally and you're a winner Take FRIEDMAN BINGO to your next briefing Never a dull moment again at CMI HINT KRYPTOS CISI CLA or any other presentation G 0 4 or More Chair No Rings in Interns Squeaks One Projection in During Asks Booth Your Row Talk Questions Man Sign A Former Microphone Snoring Language Supervisor Falls in Interpreter is in Off Back Present Slides Change without Speaker Saying Next Attendance Speaker Speaker Says uh Slide Your Book Upside Supervisor Awarded is Twice Down Walks in NOT Very Late Math-related 8 I N Phone Lights Dim Consecutively I FREE SPACE Wiring Speaker You see Sign 2 People Speaker Diagram Adjusts This Entire Asleep Pours Water on the Height of Row in Your From Screen Lectern Reserved Row Pitcher Number of People in Your Row Divisible by 3 Introducer Speaker Someone Director of Speaker Begins Leaves is is With Talk in Introduced Joke Early Attendance f Courtesy of POLEMICS 1st Issue 1989 CRYPTOLOG page 41 FOR OFFIGlAb USE O t JPY DOCID 4010026 -sEEURRE 1 r TillS DOCHMENT CONTAINS CODEWORD MA TF3RIAb -sEEURREJ---NOT RELEASABLE TO CONTRACTORS                     National Security Archive    Suite 701  Gelman Library  The George Washington University    2130 H Street  NW  Washington  D C  20037    Phone  202 994‐7000  Fax  202 994‐7005  nsarchiv@gwu edu