[This is a Mirror @ N1VUX's home page]
|
|
William Ricker's complete letter, which appears in part in the 1999 January issue of Monitoring Times. You may submit topics for posting and discussion or respond to the following letter at mteditor@grove-ent.com.
To: mteditor@grove.net
Subject: Spooky issue letters col reply
October 2, 1998
Dear Sirs,
I enjoyed the special spooky theme of the October issue. This arrived at the same time that my old jr.high interest in cryptanalysis was being reawakened by a new result in factoring (it took me a while to get the answer that Lucarelli's b-algorithm was not dangerous for public key cryptography), that I started looking at some computer security issues again for the first time in years, and that I found two cryptography books I'd bought on sale a year ago sitting in a pile. (Once is happenstance, twice is coincidence, thrice is enemy action, said Goldfinger. hi hi)
With regard to Ed's letter on a "Numbers" station that was sending only a portion of the alphabet in Morse, rather than suspecting a block-code, he should check the ADFGVX cipher and NKVD ciphers in Kahn's book or better yet in Bauer's /Decrypted Secrets/ [Springer]. (That's the one I picked up at Springer's annual Yellow sale exactly a year ago, see your local technical bookstore.)
The repeated group ATMDU between the prosigns /AR/ (end traffic) and /BT/ (break) would be the callsign of the next station being called. That callsign may be reused, or more likely was encrypted under a day-key to hide the volume of traffic going to each agent.
There are a few reasons for some of the letters not appearing (or only appearing in prosigns like /BT/) in encoded messages. First, this is probably been enciphered with numerals with letters substituted for sending. Rather than sending MORSE numerals, one uses either cut-figures or letters, they're shorter. With part-time non-specialist radio operators, such as spies or infantry, a subset of Morse characters chosen for sounding different (such as the ADFGVX cipher, where the letters are assigned from the tableau as if they were numerals) provides for better copy.
If you aren't used to cut-numerals, you might mis-transcribe them as close letters. In cut figures, the sender omits all but the first trailing DASH on 1,2,3 and all but the first trailing DOT on 6-9 . This has 1=A, 2=U, 3=V, 4=4=/SA/=/HT/, 5=5=/SI/=/HE/, 6=N, 7=G, 8=/OE/, 9=9=/ON/, 0=0=/MTM/ (or 0=T? I forget and find _no_ web reference). Some stations might send instead 1=E, 2=I, 3=S, 4=H, omitting the dashes, but I've not heard of that outside of department stores and American Morse fire-telegraphs. Of course, any station and it's operatives could have any numeral-to-letter mapping they wish. As a hypothetical, odd, but easy to remember example, by translating Roman Numerals composed of i, v and maybe x into the ternary Morse code as E,T (TT?) for 1..9={E,I,S,A,T,N,D,B,W or /BE/ or /TET/}. This maintains the roman difficulty of having no sign for 0, but for many substitution schemes, without additives, there's no need for a 0. Or they can just take a few letters that are different enough in Morse, like in the ADFGVX system, and label the encipherment tableau directly with them. Ten is more than enough, even for Cyrillic.
I would consider it doubtful that any agency using high-power shortwave would still be using simple enough methods to be cracked one-off, but considering just how easy it is to put a kilowatt or two on SW and then boost the EIRP with a beam, maybe it could be. If you have the full text, you might as well try the various statistical tests in Kahn and Bauer and see if it screams for a particular attack.
The underlying message is probably encrypted into numerals or letters in a 5x5, 6x6, 3x9 or 3x10 grid or the equivalent, possibly with straddle (meaning common letters get a single-numeral encipherment, others two), possibly with homophones (meaning those common letters each get more than one alternative numeral/s). Unless it's for a really backwards agent, it will have been superenciphered with either a transposition or a one-time-pad.
Collecting sufficient depth of traffic under the same key to enable "entry" (to make a break) is unfortunately unlikely for a hobby monitoring station, unless the cryptography is likewise unprofessional. Only if you're retired and monomaniacal would one man sit and monitor the numbers station continually. *If* their code fist is good enough, and their frequency QRM-free enough, you could put a multi-mode TNC or other code-reader and an AFC-equipped receiver on and let it accumulate on disk or in a paper pile. Of course, for a single key, monitoring the same time each day/week would likely get you things addressed to the same station -- you'll find out if the callsigns are encrypted or repeat quick enough. (E.g., if it were a simple NKVD cipher super-enciphered with a supposed running key from a book in the same language without transpositions, the zigzag attack from a probable word would work; but such would be the work of freelancers, not professionals.)
The triple UUU indicates against simple substitution but does not rule out some simple periodic systems. In order to try most attacks, you'd need a guess of a probable word, which means a guess of the language also; direction finding the station and beam width might help there, it's likely in either the language of origin or target, or English. (Sorge's ring in Japan used English for messages to Moscow, why I don't know.)
If you do break their cipher, it would, sadly, be prudent to keep it to yourself. Even if the contents are not protected by any US law on intercepted communications -- ECPA, the omnibus act, or the Lex Yardley -- you might not want the results of publishing it. If you can read it, your government surely already was and would be some put out if you published a solution that forced the opponent to (a) cancel an operation our government was set to snare or (b) replace their crypto-system or whole network. Depending who They are, they might take personal offense also; and the sort that would be unprofessional enough to take revenge might be the same sort whose field operatives and cryptographic system might be unprofessional enough for you to crack, if you could only read their language.
If you do want to take a serious crack at this, I'd strongly recommend the BAUER book, as it gives the actual techniques and math behind them in more detail than Kahn. (There are several other recent books in the technical areas that I haven't read yet, but Bauer is the one Kahn is currently recommending.) Tackling a few of the examples in Kahn, Bauer, or the textbooks might be a good practice first though. As soon as I can find my colored pencils, I want to try the 'depth' example at the front of chapter 19 in Bauer; it's in French, and the compromise for entry is given, I want to practice working it through, it's been 20+ years since I did that.
Other good books on spy-craft abound, and may be needed to understand the operational milieu; I'm not sure what's good and in print at the moment. I'm sure the books I used to like are both out of date and of print. A recent fictional book, Tass is Authorized to Announce, is an apparently well researched (supposedly a roman-a-clef) of the KGB ferriting out a Carter-era US mole who they know is target of repeated radio messages they can't decrypt -- until they raid his flat. (It's not for those who can't even consider the US government or its agencies doing anything mean or illicit.) Puzzle Palace and Deep Black both focus more on the current technology and less on routine field-work spy-craft and its cryptographic protection.. There are internet discussion groups at news:sci.crypt (?and news:alt.crypt?) where discussion of such projects might prove fruitful.
Spookily yours,
Bill Ricker <wdr@world.std.com>
ARS N1VUX / BOSTON
Return to the Monitoring Times home page
[This is a Mirror @ N1VUX's home page]