o 11e• 1 W UOl all1 il lDWOU lBl WCB0' Ul llllV lB li11 ls WB'1ll Jls f lilullllf wml l I PL 86-36 50 USC 3605 I I • I I 1• I - - ' r j 'u -m - ' · -------Ti--- · - ·······l T I P S_I S S T l LL A L l V E hN D W E LL • • · · • •· · • • • • · z • ··· • •·••3 •• • 6 •• • ••••11 9 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ·_·____ ______ 12 1 _••••••••••••• •••• • ••••• 14 TRANSLATOR I L 15 MORE COMMENTS ON THE·AG-22 IATS OUESTION I SEMANTIC VOIDS DON'T SHOOT THE MACHINE-PRODUCED AIDf fllll 111•1 1 • 1 IIFPlldl • 1111ft 9 • 1 111 rl-• eP llfi• II - a llMit DlllNSA CHCURlC'° ' nmptr -p I ii 2 p I t U eCllcalllab11MOrlplatar Declassified and Appro ved f o r Release by NSA on 04-13-2021 pursuant to E O 13526 MDR Case #109435 Ff9P iJDOllliFf Published Monthly by Pl Techniques and Standards for the Personnel of Operations VOL III No 9 SEPTEMBER 1976 I Non - Responsive i - -- -· -------·---- '£ ' - - ·- 'W' ------- ·- · - - - _ @10011Hlf JfJ iA I PL 86-36 50 • TIPS is a part of RYE a UNIVAC 494 remote-access machine processing system at NSA The TIPS system encompasses Bu Ue tln Vol 3 No 9 I s 6 ue No 30 the hardware devices software executive Sp ung 1976 l 6 11 ep ti nte d heJr e r ulth the routines conventions communications luttd peJUnl4 s ic n 06 C-LINERS Ec U to11 package and data bases in support of Vavid 1 WUliaJ 16 the quick-turnaround on-line information storage and retrieval capability within RYE See Section 4 of forthAs was the case with Mark Twain reports coming USSI0 703 Technical Information you might have heard about the demise of TIPS Processing System TIPS for general -- NSA' s Technical Information Processing System information about this system -- have undoubtedly been grossly exaggerated Under the TIPS concent the data files are True TIPS is showing signs of aging After stored on CDC-606 tape drives They are conall she has been around since the mid-Sixties nected to Honeywell DDP-Sl6 mini-computers To some such longevity should qualify her for which execute the queries submitted by the varisome kind of geriatric support To others ous users A query is a series of simple statenotably some folks in cil the Information Sys- ments written in TILE TIPS Interrogation Lantems Division of C the old girl is still very guage Normally it might consist of just a much alive Admittedly a young and more glam- retrieval command and a display statement orous replacement is being sought Nobody knows Queries are entered remotely from a teletype when this rival will be embodied ·cor em-machined terminal or some other peripheral device like but she is coming and plans are being made for a Raytheon CRT or a Bostic paper-tape reader her arrival Chances are that as a RYE user you've At this stage in her career then one feels already interfaced with the most cormnon of it would be a good time to record a few random these input devices the lowly teletype The thoughts about TIPS A few words of background manufacturer is the Teletype Corporation of information may be in order for those not in the category of C Old-timers First a more or America and the most common terminal type for RYE is the ASR Automatic Send and Receive less official definition The 60-U ow ing a 1 ti cle bt C-LJNERS C GJteup Machhte PII OC Ul i ng Zn60J1111a-UOn September 76 CRYPT0I OG • Page 3 SSCPlilil Iii PlPi 5 iii I 22 i rr IF 3 1 F L use 3605 I 81QADlf and-forward switch housed in a PDP-11 main frame and located at the DIA In turn COINS hands Incidentally you might well impress some of the data systems newcomers at the Agency interfaces with the so-called IDHS Intelligence Data Handling System which links a number of by reminding them that all teletypes are teleprinters but not all teleprinters are teletypes remote customers indirectly to TIPS the NSA -- especially if you've heard them talking about system These include xeroxing their output with a small x I • Air Force Intelligence AFIN Pentagon ASR means you can input a paper tape in • Naval Ocean Systems Center NOSIC the reader while at the same time receiving Suitland Maryland and page-print at 72 characters a line The read• FICEURLANi Norfol Virginia ing speed certainly isn't the fastest by today's They also include such very re1110te customers as standards -- 10 characters per second or about • Air Defense Command AOC Colorado Springs 100 WPM By contrast the Bostic units each o European Command USEUCOM Vaihingen controlled by an ASR-35 terminal can input at Germany and a rate of about 300 characters per second and • organizations subordinate to the Pacific punch paper tape at a speed of 105 characters Command PACOM such as per second • PACAF Hickam AFB Hawaii The Mod-3Ss are flexible however and easy • ClNCPAC IPAC HQ Camp Smith Hawaii to operate There are nearly 100 of them in • COMUS KOREA Yongsan Korea operational NSA spaces including FANX hooked Now active on the TIPS syste11 are about 40 into RYE This figure doesn't include the separate SlGINT files not including about 1S units belonging to C system organizations Note al o that it doesn't include the consider- support files for accowiting user-aid and test purposes They are managed bv owners snread able number of Mod-3Ss connected to TIDE Time across A B C G P V w 1 Dependent System housed in a similar U-494 main frame The breakdown for ownership of the RYE-connected units is s OPf s • _____________ • Model 35 or simply Mod 35 to the old RYE I I El • • • • • • • • • • A 8 -_ Cl CS E 24 G 2 R P 2 V - 2 2 8 10 W - 15 12 - Some of these 40 TIPS files are on the 13 - technical side Admittedly they are designed more for the SIGJNT nroducer than - 1 _ __ Thi aQ_ove list doesn't incl ude the non-NSA • service intalligence organizations i e • AFINAR Air Fo Cf Intelligence and Research • USASRD U S A Sp cJ al Research Detachment and NFOIO Naval Field-Qperations Intelligence • Office each of which hits a terminal and can • access selected TIPS COINS litqs • SIGINT users like the • ------ 11 and the USAFSS at Kelly AFB Texas ave erminals connected to TIPS They operate like regular TIPS customers i e they are linked directly to the RYE 111aster machine and from there to the TIPS data bases to which the have been iven access 1 The second major category of TIPS files could be termed the intelligence type _ These are files available to both NSA users and the intelligence conanunity through COINS One big subset of the intelli11ence files comnri s- • • 11 Incidentally if you have come aboard the Agency fairly recently you may be confused about the relationship of TIPS to COINS TIPS i the NSA mode of the Community On-Line Intel11 aence System COINS -- a network of intelligence-couaunity computers which have been in place since 1967 in either pilot-experimental or final-operational mode COINS currently int'lrfaces with the NSA system through a storeEO 3 3b 3 EO 3 3b 6 PL 86-36 50 USC 3605 ---------- The TIDE file carries the latest 20 days of September 76 • CRYPTOLOG Page 4 SFCRIJlf EO 3 3b 3 PL 86-36 50 USC 3605 liillilSRllf EO 3 3b 3 PL 86-36 50 USC 3605 ip£ercept the TIPS files about 90 days of data 21 or more days old Another recurring problem has to do 'with the terminal equipment itself Like all of us the Model-3Ss are aging at the same time they are occasionally a prey to gremlins and mysterious poltergeists Maintenance service however is normally prompt Call RYE operations S93Ss to enter a complaint and arrange for servicing Do you need a beginning or refresher course in TILE TIPS Interrogation Language Formal training is offered aperiodically by the National Cryptologic School MP-17S If you are interested in such a class call E21 8555s or Cll The COINS Project Management Office does a good deal of informal training in TILE as well guages used to access COINS files at DIA through its user support 1 0 JP'- ou can reach them at 1108s Cll3 can also arrange informal training briefings for potential users or analysts who would like to have a refresher in any aspects of TILE At this writing the long-tem future of TIPS is hazy We are still in the era of TIPS I of SIRE the SIGINT Requirements Will there be a TIPS II No one is saying and • file managed by Vl is off- loaded to TIPS for the benefit primarily of remote SIGINT producers presumably a final decision has not been made At the moment one assumes that the TIPS II maand collectors SIRE is actually maintained chine or whatever system replaces her will be through the SOLIS system strongly interactive - - with all the features and high overhead that this capability entails But one hopes however that the time-honored batch 110de will not be scrapped completely A lot of us old-timers and some younger ones who have grappled with day-by-day general processing at NSA like to point out a simple but noteworthy fact the remote-processing batch-mode systems of yesterday are still here and still performing prodigies of labor for the Agency At this writing t e vo ume o a a on e TIPS system totals about 300 million characters One of these is SPECOL Special Customer Oriented This volume is lllOre tnan we once thought could be acconnodated If you are a potential owner of a new TIPS application this shouldn'tneces- · sarily discourage you unless you are thinking lh u- ex- r e me- y e- ec 'P v e l'l ll'l ' n o rm a t 1 o n S t o r a g e• of submitting a large file for consideration • • and Retrieval and data-processing system for A number of current files are scheduled to be • several different machines including the 360 370 removed from the system This should free so111e world RYE TIPS is another workhorse limited space for future applications Call Cll for to a UNIVAC main frame but reaching out as far 110re infol'lllation about available space as Europe and the Pacific to perform its IS R To many potential TIPS COINS users at ij A and data-processing role Note that both SPECOL a perennial problem has been ''Where can•I find and TIPS are 1110re than IS R syste111S They are a convenient RYE outstation from which lo input data-processors as well as retrievers able to my queries Terminals are apt to be iocated extract sort compute format and output inmostly in machine rooms that may or mly fo1'1118tion in many different ways and for many not be handy to your office Admittedly this different kinds of users Can you say the same is far from the ideal situation Pa rt of the for YOW' interactive system problem is that Mod-35s tend to bJ a little These random thoughts have not been aimed at noisy and not conducive to a seNne office atdisparaging the young DBM systems we are now mosphere Perhaps one solution•would be 8 ogling Assuredly they have a bright future at telephone call to your friendly machine support NSA although their outlines are a bit 111Urky unit e g G8 for G f i les or B42 for B appli• yet But let•s not be in too big a hurry to cations which could make i'he query for you divorce ourselves from the old batch-1110de sysIn any event Clll S203s which functions as tems that have served us so well for a long a TIPS customer support ftit in addition to ti•e- They deserve at least a glass or two staffing the NSA COINS Subsystem Manager's Ofraised in tribute £ice can guide you to ard the nearest outstation or uke a qu ry for you • c• •etsi ----------- EO 3 3b 3 EO 3 3b 6 PL 86- 36 50 USC 3605 September 76 CRYPTOLOG • Page 5 8B@RDf PL 86-36 50 USC 3605 I • IU l Br l d Sttee Jto -e eo fft e11t41 t tie j-22 1 17'S J- e4tio1e l I _ _ _ _ _ _ _ _ _ I - 09 W wJ Cecil Phillips' article Musings About the AG-22 IATS in issue No 28 of C-LINERS and reprinted in CRYPTOLOG March 1976 caught my attention because of the recent work I have done on data bases built from AG-22 IATS and not too long ago FF STRUM I do not think that anyone would argue that computer records built totally automatically from AG-22 and IATS are of low quality because of poor planing or a lack of sophisticated programming The quality of the copy itself is obviously to blame and I do agree with Mr Phillips that the place to solve the problem is at the source· however I disagree that 1110re operator tags ' and less traffic is the solution I cannot recall any change or addendum to Morse copy instruction within the last 20 years aimed at lessening the amount of non ntercept data an operator lllUSt manually enter into his log traffic service In fact the trend has been in the opposite direction The implementation of AG-22 brought tagging· diminishing intercept resources deunded that we narrow our requirements and specifi cally define the collective objectives and now COPES i s with us The Morse operator is and always has been overburdened and it seems t o be some accepted rule of collect ion strategy that a position must be assigned more cases than it is humanly possible to cover There is something drastically wrong with this rationale Most of our Morse operators from the Crypto• logic Services are placed on the job on OJT just out of school and complete their military enlistments with less than 24 months on pos i • tion There is absolutely no comparison between U S copy from a quality standpoint and that of second- and third-party sources where genuine professionals are universally employed I'm not trying to put down our operators· on the contrary I think they do an excepti nal job with the limited amount of experience they have _on the average but leaving i 11portant lineby-l1ne copy decisions i e summarizations up to them as Mr Phillips has suggested is not at a ll practical Weak reception highspeed transmissions signals buried in QRM coverage of both ends etc require intense concentration during copy by the operator In• tercept of transmissions under such adverse conditions is COllllllOn and would hardly lend itself to a form of selective abbreviated copy Gi 11tina that which is ihe usual also requires that the experienced traffic analyst or crypt -Re_plli t t ___ ed _61LO_m_C___ J IN_E_RS_ Vol 3 No 9 Sp1W19 1916 I uu e No 30 analyst do it especially if enciphered procedures are used There are a number of important problems in this Agency today where non-11essage data is all that is readable Messages are either high grade unsolved or practice and hold little or no analytic value Copy instructions for certain groups passing only practice sometimes require only the first line or so since we anticipate these texts will be repetitions from recovered pages The chatter whether the messages are of value or not may contain routing instructions frequency and schedule references authentication message precedence etc each of which is essential in determining communications procedure i dentifications and most important network continuity We have even had callsigns selected in a particular order from a recovered page used as time indicators in missile-launch countdowns Order of callup has also been used to establish continuities International Q and z codes have been used for purposes other than originally intended There are iaany examples of Q's and Z's being used to indicate precedence and authentication or other uses peculiar to a target's need Abbreviated plaintext chatter prevalent on many problems also requires special attention Full copy of chatter less the strings of V' s repeated calls etc is important to any analysis to be perfomed on the material Analysis of the collected data f luctuates according to need When continuiti es are good and callsigns projectable less emphas is is placed on the study of message externals There are many traffic analysts here today who have never worked on a major target complex where the callsign system in use was unknown Traditional TA for some of them i s an unknown art that will be painfully rediscovered when these callsign systems change We have enough trouble wi th routine changes within a recovery system so let's not lose the remaining means for reestablishing the continuities when callsigns can no longer be used as the primary lead to an identif i cation or continuity There is a great deal more to TA than callsign l ookup and it lies in the detailed study of all co11111unications data especially in the chatter We should not omit or attempt t o summarize this kind of information at the point of copy Extra intermedi ate edit steps to fi x the traffi c copy are no solution either Manual September 76 • CRYPTOLOG Page 12 81tJIIM' MOWi S El I am 111 TI CPI PIP sr 2 ii S6981fr intervention to correct an automated process seems to me to be a step in the wrong direction We will always need people to prepare the input operators and others to study the output analysts Mechanization should join these two functions if we need to fix the system let's do it on either or both these ends not in the middle Middlemen editing and or correcting ' large volwnes of traffic with strict deadlines to be kept contibute very little to the process since we are frequently forced to accept anyone from any specialty who happens to be available to do this job Re-identification except on search material as part of a maintenance process performed back here perpetuates rather than corrects a bad practice Wh_e intercept identifications must be changed this means that the wrong target was copied as mission It is our responsibility to see to it that the operator gets the right information to acquire and identify his target properly in the first place After-the-fact case corrections are useless to him if the SOls i e callsign periods are short in duration Any other type of correction alters the original copy which should not be done for a multitude of reasons File maintenance if it's worth doing also requires quality control which ls turn requires time and resources The value of manually correcting traffic for data-base storage is questionable since the case analyst is usually through with it at the close of the work day whether it is properly formatted or identified anyway He has logged the important items from his material updated his net diagrams etc and will probably never need the traffic again for the types of specific work he is required to do e g TEXTA in one of its many forms which is in turn machined I think it is best that we continue to get the best possible operator copy and retain all of it and the identifications his any intermediate machine re-idents and the final that are placed on the stored copy Under no circumstances should we change the original version of copy or swmnarize any part of the data that does not follow a predictable pattern such as valid messages and chatter even if the instructions SCOLs etc require no 1110re traffic than is necessary for identification If we need to de-dupe let the programming handle it but I would be hard put to support my conclusion I suspect that I would have no case at all if I had to provide supporting evidence based on past usage of the data bases or the applications of special research beyond case analysis which is often sacrified in favor of continuing projects and the fulfillment of day-to-day conunit ments The potential uses of these data bases through readily available programs should bolster the TA imagination and create new approaches if properly publicized and if adequate indoctrination is provided at least enough to get us beyond the customary case and data retrieval with follow-on sort and list that is usually requested by the average analyst In spite of the more obvious shortcomings of formatting and processing traffic for these data bases I feel they are valuable whether corrected or not and that we should promote the use of the numerous facilities and software that we have at hand in exploiting them Getting this done is another gigantic problem in itself that needs the attention of management in particular The problem of formatting traffic for database storage is an old one with quite a history I can remember my efforts some 15 years or so ago to have field analysts edit traffic for electrical transmission rather than prepare the detailed complicated TECSUMs MATSUMs of the day The idea didn't make it but the same general concept of editing for data-base foraiatting I tried to sell is now the respons_ibility of the AG-22 operator called tagging The formatting of Morse traffic has been tied to a number of vehicles over a long period of time and evolution from TECSUMs to MATSUMs to STRUM to AG-22 All of these have had a l leasure of success but now that we have managed to automate forwarding directly from the intercept position our concentration should still be on intercept and the improvement of copy 1 believe that placing the formatting responsibility almost totally on the operator through tagging will divert our collection from these objectives and generally degrade traffic q lity We should be able to do without some of it or substitute 111Ultifunctional and or automatic tags and develop programs with the objective of doing a better job of data-base formatting that are a little less operator-dependent So this brings about another question If the case analyst doesn't need a traffic data base in the normal conduct of his daily duties who does Management sometimes for collection studies from data not available anywhere else but this certainly does not justify a data base The real user is the research analyst studying larger masses of both identified and unidentified traffic Here is where new ideas and analytic concepts are born It is for this reason and perhaps this reason alone that we maintain a nwnber of these data bases I think we are justified in doing this for just this purpose Surely we have the expertise on hand to do this We have a pretty good system work_ing for us now not just in AG-22 IATS but in a variety of other areas e g automated callsign projections machine decryptions radio wave propagation cryptanalytic diagnostics etc If we c an teach the lllllchines to do these things there is no reason why we can't get on with the task of formatting traffic acceptably with follow-on programs doing the specialized work I do know that edit programs have been written to fix message tests from existing data bases not only for decryption but also for indexing September 76 CRYPTOLOG Page 13 8BftllM I 81611ii1 diagnostics etc as well and the crYPtanalyst has been generally satisfied with the product There have been programs developed that scan chatter find enciphered address groups decrYPt them and place them in sort field key locations so they can be listed in an orderly fashion and in context We should be able to expand on these techniques and do a fairly good job of mechanizing the editing and forinatting of AG-22 IATS take with minimal tagging I contend that if we retain selective retrievability in these data bases so that we can get back to this material through identifications case and terminal for the specialized processing that needs to be done that is sufficient Successful traffic analysis depends upon traffic quality Let's do something that will assist the operator in making that product more accurate complete and at the same time 111ake · his job easier It's time the machines were put to work to serve us @151 Z September 76 • CRYPTOLOG • Page 14 IJl@RI •1 Non - Responsive 11 rwr a z· n 7 IL EO 3 3b 3 EO 3 3b 6 PL 86-36 50 USC 3605 8DORlf 1I '• i I 1 September 76 • CRYPTOLOG Page 15 fltt SIE 1171 7211519 J J mu DI L EO 3 3b 3 EO 3 3b 6 lillODllfl PL 86-36 50 USC 3605 September 76 CRYPTOLOG Page 16 liHiJiBlllfl um 5 III I nrn• Ill 5 C Si Zi llQP iililHiFII ur 1111 'l OS The COMINT linguist has at his disposal in addition to general and specialized dictionaries various types of language working aids that arrange individual words or connected text in ways that look unusual to the linguist 1 • The purpose of this article is to acquaint the nonlinguist and the linguist with no previous COMINT experience with the COMINT need for those working aids and with some of their characteristic uses and limitations But before discussing these aids it might be worthwhile to e7 plain why lan'11age aids have been created to manipulate words in ways that seem to differ o_greatly from the normal patterns of spoken and written language Unlike Moliere' s bourgeois gentleman who was surprised to learn that he had been speaking prose all his life there are probably few who are surprised to learn that normal spoken and written language is unidirectional That is people start speak·ing and whether or not they have previously organized their thoughts logically they produce their words one after the other in a definite irreversible time sequence Or they start writing and depending upon the particular language put the words down in a definite sequen ce from left to right right to left top to bottom etc This text for example has the words arranged the way we like them -- and it's not just because a typewriter couldn't be hooked up to type the words boustrophedonically • PL 86-36 50 The listener or reader perceives each string of words in its produced sequence and then is able of course to scramble them at will in his mind He does not usually listen or read by jumbling the words back and forth out of time or space sequence Nor does he usually listen or read backwards Spoken or written sequences that are actually intended to be interpreted in a backward sense are usually contrived for artistic or comic effect Such contrived sequences include palindromes Able was I ere I saw Elba typographic tricks caption on a cartoon showing two people shouting to one another as their respective cable cars pass I 1aid wait for me at the other end and bn• edto •ilt ID •m 10 ti aw b iaa I or dialogue in animated cartoons the hero propelled at breakneak speed over a hazardous course stops dead to exolaim What a buggy ride continues on his way until he crashes and then is propelled backwards over the same course during which he stops to exclaim Ride buggy a whatl Except when creating or responding to such stylistic tricks normal people that is those• not suffering from some psychological or physiological impairment of the ability to produce or to perceive speech or writing handle lan11 Ua2e · strin s in their usual sequence But COMINT Ian- guage specialists do not deal with language data that is normal In the everyday world a person on the end of a telephone line who does not hear a word perfectly can ask the person at the other end to repeat it The NSA transcriber obviously cannot do so In the everyday world an interpreter seeing a confused expression on the face of the person for whom he is interpret_ing can clarify the linguistic ambiguity right September 76 CRYPTOLOG Page 17 use 3605 I bi Sl l@IRRf lfftlllllf EO 3 3b 3 PL 8 6-36 50 USC 3605 FIOP 9lilfiJ8ilF lllf RP ft l then and there The W A_• l inguists has to deal with the ambiguities as tliey•s tand He said June but I think he ml ant' July In the h everyday world a person who receiv I an importThe NSA linguist's job is analogous to t at ant telegram with a cr tical word garhl'eQ can of the FBI specialist in the chemical-analysis• or ballistics laboratory But whereas the test request a repeat The SA anal ' t can onry • run by the FBI specialist weigh and measure the hope that the intended recipient wiU be just as • • confused as he is and will request •a repeat guantitative properties of physical objects • that is also made availdble to the NS i analyst iho e run by the NSA linguist assess the quali- If he does not the NSA li ist has to de garble tatHra l roperties of words And words are things • the text as best he can even if that in•olves that keep hanging If an NSA linguist has a • his going out on a limb • • garbled S-ldtter word how does he decide which · of the three obviops degarbles is the most ikely NSA linguists accustqpled to dealing with• to have been the wo rd intended in the original these and related proble111 that arise when Ii - text If an SA bookb re ker has an unrecovered• tening in on foreigners• eonversations and • value in a one-part code be1 11een letter Mand • reading their foreign-langtiage telegrams know ' how does he decide which are the pest guesses that often they cannot atta k a foreign-language r an NS Tfl arch analyst has a 111e sage signed text in strict left-to-rigllt o rder Nonlinbyl bow does he deci_de which I guists sometimes ha e the i111pression that just it is Or if he has in his traffic an abbre-• like a kid who can't have lfis dessert until he viation with 40 possible expansions how does • has finished his carrots an1 string beans the he determine which is the one that the message • NSA linguist has some profe1sional or moral oboriginator had in mind and that the message re- ligation not to look at the second word until cipient will recognize immediately Inciden- • he has translated the fi rst -one or to look at tally how in dealing with intercept text writ the third until he has trans'lated the second ten all in capital letters did the NSA linguist • etc But this is not true eyen in what might recognize the abbreviation in the first place be called normal translation work when for The answer to all these questions is the same • exa111ple a co111111ercial ' that i non-COMJNT he made a judgment based on his thorough know- • technical translator is translating a completely ledge of the language with which he is dealing • ungarbled text fro1D a printedi open-source book not just a thorough book-l'arnin' knowledge • or magazine -- when translatif i a technically of the language but a thorough knowledge of it complex or grammatically obscui-e sentence he as it is actually used by the specific user may indeed have to look ahead to the next senwith al 1 his speci fie educational occupational • tence to the next page or ev n to the next social and other peculiarities and based on chapter for clues that will r solve the ambigu• his submitting of the language data to validity• ities And the situation is evfn more complex tests at all stages of intercept decryption for the NSA linguist who often finds that betranslation and interpretation These tests cause of message encryption ga bling poor that the NSA linguist subjects his language audio signal etc it is impossibl e to attack data to are just as valid a s the chemical and a written text in a strict left-to' right sequence 2 spectrographic tests conducted by physical sci - • or to transcribe a radiotelephon conversation entists but the results of the linguistic anal- from the first syllable to the la st on the tape ysis do not have a nice solid scientific look to them A typewriter expert can state in • The NSA linguist is a kind of scientist in the court ''The letter m in the examined document • sense that he can isolate the words he wants to could not have been ll de by a Slllith-Corona type-• • exa mine and can subject them to any k nd of test writer and stand ready to back up his statement • he wants ter to ext ract their intelli 2ence with enlarged photogl aphs showing measurable • • ••a I •ll-·-----N distinctive features But the NSA linguist who • 11 • states with identical firmness ''The letter m 1 • I •i n this word must be a garble for a usually ca6 Jiot support his findings as impressively and might even sound downright shifty as he brings in sucll qualifiers as usually cannot OT refers --•l owover modestly -- to his years of experience feel for the language or to l ett er-frd luency pl Obab i lities contextual incongruity • and other qualitative rathel t ban • quantitative pi-oofs How then do the NSA linguist equip himself when attacl ng COMINT text in a particu- • lar foreign language' Obviously the first sl'li p is to acquire new and newer dictionaries 2see Right-to-Left Text Sorts Are Not and to•iu apent them witltpperational language Impossible by CRYPTOLOG files and 9PVCialiied glosaaries But these all list words in_ he noma1 alphabetic order August 1974 • • • •• • ---------------------- -1 I I September 76 CRYPTOLoG • · Raae 8 'JIQP ilti Rllff IO IIR1fJ • •••_• -P_L_8_6__3 6_ _5_0_ U_S_C_ 3_6_0_5- PL 86-36 50 use 3605 TAP SEGRlii' llJ IDR1l 't _ _ _ _ ____________________________ and are therefore of J imited value to the crypt- analyst or linguist t s ing to cope with a word with its beginning missing garbled or cryptographically unrecovered 'J'h•refore NSA analysts speciali2 ing in language' uwe felt the need to prepare VIU'ious types of l anguage aids that list words in other than norival dictionary orderl The aids fall gen al1y into three basic categories • • • word-pattern listings • backward listing • • window indexes • Word-Pattern Listings Word- pattern listin gs are•listings usually of preselected words Oo•eti es including word phr LSes of two or thre words that are typical of a particular type of contex t and that have a particular pattern of letter•repetition For example the English wotds AARDVARK EEL LLAMA CANNOT and WILL each ontain a doubled- letter sequence that can be rcil resented by the coding AA The words LLAMA COMPLETE and MILITARY each contain a repeated- letter wi th one intervening letter which pa tern can e represented by the coding A-A The phrases CAJ 1 NOW and TO OPEN each contain a doupled-letter sequence that can be represented either by the coding AA if the space between words•is nonsignilicant or A-A if the space betwe n words is gnificant II ' September 76 CRYPTOLOG • Page 19 iror MIOIIE• tJPIIBftil EO 3 3b 3 PL 86- 36 50 USC 3605 EO 3 3b 3 PL 86-36 50 USC 3605 1• 1JOP QIHJRll• l fllBllfl C I _ _ J _ • ______• 6122 ssnr rzr· To be concluded next month September 76 CRYPTOLOG • I Non - Responsive - ••••••••• Page 20 or @l l@JRlf 19PSIB I ' •• •• • · • 1• • • • • • • • I • PL 86-36 50 USC 3605 I