Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
Encryption Security

Secure Digital Voice Communications In World War II 34

Posted by michael from the good-use-for-a-phonograph dept.
mercury7 writes: "Saw this one on Memepool. A very interesting paper from the U.S. National Security Agency site on the first digital encyrpted voice communication system. It is incredible how hard it was to manipulate data before the existence of computers."
This discussion has been archived. No new comments can be posted.

Secure Digital Voice Communications in World War I More

Secure Digital Voice Communications in World War I

Comments Filter:
  • Why is the NSA publishing this? You know, reading the first nine tenths of that piece (and the slashdot one-line summary), one might think that the U.S. Intelligence community (the NSA wasn't around, right?) pulled off quite a coup, developing revolutionary, ahead-of-its-time-digital-telephony-technology. But the appendix buried at the end comes to quite a different conclusion based on the testimony of someone responsible for the project at the time:

    "All of the elements of this system were developed by the Bell Telephone Laboratories in the interests of advancing the art of telephony. . . . When it had reached the point where its principles could be demonstrated, the Signal Corps took prompt steps for its procurement."

    It appears to me that the NSA has, to a limited degree, tried to make their mammoth budgets and spare-no-expense approach more palatable to the American taxpayer by claiming to develop technology way ahead of its time. "Hey, our job is American security but even if you think that's not a big problem these days, look at the tech spinoffs and innovation we create." To me this seems a little disingenuous.

    As this article, among others indicates, the NSA et al may be a nice mechanism *funding* innovation (although VCs are arguably better, with a more attractive upside economically) they don't actually *do* much innovation (outside of say, crypto research).

    --LP

  • Re:Key distribution on vinyl disks! (Score:3)

    by Detritus ( 11846 ) on Saturday January 13, 2001 @06:00AM (#510266) Homepage
    You can listen to examples of vocoder audio on this web page [berkeley.edu].
  • Also, they seem to think they can pass secret instructions to spies the world over by inserting 'random' bold tags on one of their pages

    That's because of the GCHQ challenge - try it here [gchq.gov.uk]!

  • That's very cool; I'm amazed they understood enough about the importance of using something truly random to generate the 'noise' for the key records. Digitally, I wonder how much data can be stored on an LP (were they 78's, 33's, 45's - see the NSA still won't tell you everything). I wonder how they transported the key records... it's a whole Bond movie - the Mata Hari type seduction attempt to get the record from the young Signal core engineer on the PanAm flight. Menancing German agents sneaking into his hotel room to record the record.. Oh well, good post.

  • While this is certainly funny, and I wish I could mod it up further, the original first use of a vocoder in music was done by the Alan Parsons Project around 1976 or so (does anyone remember the album/track - I can't remember if it was Stereotomy, or Tales/Raven)...

    Worldcom [worldcom.com] - Generation Duh!
  • "Two turntables and a microphooooone..."
  • It appears to me that the NSA has, to a limited degree, tried to make their mammoth budgets and spare-no-expense approach more palatable to the American taxpayer by claiming to develop technology way ahead of its time.

    You might try doing some research before making such accusations. Yes, the NSA takes advantage of technology developed outside the agency, however, the NSA and its predecessors have a long and distinguished history of research and development in communications security and intelligence. The problem is that most of it is classified, so you may have to wait 50 years to learn about it. They still haven't declassified all the material from World War II.

  • What they have done is a real One Time Pad on voice. By quantizing the voice and combining each sample with an equal amount of entropy modulo the sample size (in this case 2.5 bits since there were only 6 possible levels) you get perfect secrecy. If you assume that the key is completely unpredictable then every possible outcome is equally as likely. They can only do this by breaking down the voice into samples of a known size. Check Claude Shannon's groundbreaking paper "Communications Theory of Secrecy Systems" for more detalis.

    Burris

  • "It is incredible how hard it was to manipulate data before the existence of computers."

    Yes, back then they couldn't just run it through a computer and have it scramble it. To do this they would have to run it through some sort of mechanical device.
    --------------------------------------
    I'm a karma whore, mod me up damn you!
  • where the reciever broadcast 'noise' at the same frequency as the incoming message. As the reciver knew exactly what noise he was broadcasting he could then subtract that quite easily. So you don't NEED digital encryption.

    Obviously this would only be effective on point-point (cable) connections, as the noise would become insignificant when broadcast by an arial.
  • Claude Shannon wrote his seminal paper "A Methematical Theory of Cryptography" just after the war in 1945, though it was classified. This is the paper where he proves that a Vernam cipher with a truly random key (a.k.a. One Time Pad) provides perfect secrecy. It is well known that Shannon was the main man for US crypto during the war. So it shouldn't be a surprise that they knew exactly what they were doing.

    I very highly recommend everyone read the paper, which was published in the Bell Systems Technical Journal under the title "Communication Theory of Secrecy Systems" in 1949. It's pretty accessible, you don't need to be a crypto weenie to understand it. However, you can't even pretend to know anything about cryptology without having read that paper.

    Shannon is the MAN.

    Burris

  • Sounds like one further step from MPEG layer 3 and MP4. M$, Thomas, (and everal other corps), take that!

    At least, it'll work for voice. (If I understood this post..)
  • Once I read something about the US, using native americans in special tasks, they could just speak their natural language without any encryption, because, anyway no one in the world could understand what they said!


    Yep. Someone aluded to that earlier. The Navajo Indian tribe spoke a language which was so far removed from any other dialect then in use by any other group in the world, including neighboring Indian Tribes, that it would be impossible to decipher without some kind of visual reference. Couple that with the fact that the language was isolated and nearly dead, and it made the perfect language for encryption. Navajo Code Talkers were assigned to nearly every unit in the Pacific Theater, and were in the first wave of nearly every assault. Just as the Nissei or Buffalo Soldier, the Navajo code talkers were a crucial part of our military operations worldwide, while their people back in the states face brutal oppression on many fronts. That says a lot about someone who so loves their homeland that they'd die to defend the neighbors that hate them.

    *sniff*.
  • It was Wendy Carlos who first had the idea of using a vocoder for music, in 1964. See http://www.wendycarlos.com/vocoders.html
  • Ok, there's probably something here that I'm not getting, and maybe someone can explain it to me. But after reading their description, I don't really understand why they went to the trouble to digitize the signal at all.

    If you're going to have to produce and distribute "noise key" records in the first place, why not simply combine the two signals (the voice and the noise) in an analog fashion before transmission, and then do the same process in reverse at the other end? This would have been MUCH simpler (meaning they could have deployed it sooner).

    Granted, they get an extra degree of security by digitizing the signal, simply because the enenmy must then reverse engineer all the digitzation hardware. So rather than simply needing to steal a copy of the noise key, the Axis would have had to steal 55 tons of equipment to use it. Still, I don't see how the system is made fundementally more secure by digitizing the signal. And I'm sure they were in a big hurry to get this system in place, so I can't see them making it more complicated than they absolutely had to.

    Can anybody explain this to me?

  • encryption, etc (Score:3)

    by Alien54 ( 180860 ) on Friday January 12, 2001 @09:18PM (#510280) Journal
    In many communications of this type, the aim is to try to make could be intelligible speech indistinguishable from noise.

    Note that many politicians and spin doctor seem to have this done to a fine art.

    The reverse is slightly more difficult, which is the fine art of taking something indistinguishable from noise, and try to extract intelligible speech from it.

    How similar this is to public forums such as slashdot, etc is left as a exercise for the reader.

    ;-)

  • Ok, there's probably something here that I'm not getting, and maybe someone can explain it to me. But after reading their description, I don't really understand why they went to the trouble to digitize the signal at all. I wondered that, too. I got the impression, though, that they wanted the signal to be undetectable in the relatively-broadband HF receivers of the day, not just undecodable. That meant they needed to distribute the frequency content over a larger chunk of the spectrum, making it sound more like random noise in the AF bandwidth of a conventional monitoring receiver. Wideband linear FM modulation would have been a real hassle in those days (no varactor diodes, for instance), so it may actually have been more convenient for them to drive six separate transmitters with the results of the "digitization." OTOH I'm pretty sure that contemporary telephone gear was already using frequency-division multiplexing without any digitization at all, so I don't see why they couldn't have accomplished the same thing with an analog filter bank and mixdown scheme.
  • I thought that part of cryptonomicon was speculation on Stephens' part. It's cool to find out that this was not only real but still documented.

    What a clever idea.

    -Peter
  • Hello,

    being a native german, I have read lots of books about WW-II, as I would really like to know how this all could happen, I was lucky to get some info from my father, who was a soldier during that war and at the end POW of the U.S. army, a bit irritating, but from the things he told me, that seemed to be the best time he had during the war and afterwards during his life, he was always a friend of every decision made by american politicans. Sad, that I can't ask him anymore any questions...:-(

    Once I read something about the US, using native americans in special tasks, they could just speak their natural language without any encryption, because, anyway no one in the world could understand what they said!

    If someone has more info about that, or even an URL, I would be very thankfull...

    Michael

    I hope it was not a mistake to click on this NSA link, but I still can't hear/see any black helicopters above me...
  • I knew it. We need to criminalise the production and possession of vinyl as a munition. :-)

    The technical hurdles they had to overcome for this first "digital" voice system were pretty impressive. And each station weighed in at a mere 55 tons. I'd love to hear a recording of what the recovered speech sounded like.

    the AC

  • Computers did exist! (Score:3)

    by the red pen ( 3138 ) on Saturday January 13, 2001 @08:03AM (#510285)
    This is a trivia sidebar.

    Another thing that was tough in this age was aiming artillery. If you know the distance, the characteristic of the gun, the characteristics of the shell, and the windspeed, physics equations tell you how to aim the gun to hit your target. Realistically, the equations cannot be solved in battlefield conditions where time is a factor and competent mathematicians are in short supply.

    The solution to this problem was to issue huge books of pre-computed tables to artillery gunners. Each new gun and new shell needed a new book generated. That involved many hours of computation to be done by a small army of geeks. Because the guy-geeks were busy cracking codes, inventing atomic weapons, running logistics and stopping lead, a cadre of girl geeks was assembled in Annapolis and tasked with generating these artillery tables. The women were brought up to speed on Calculus and Physics (if they needed to be) and they were know as (you guessed it) "Computers."

    While the lady Computers toiled, some other geeks were trying to automate their task using analog and digital electronic systems. Because the first automated calculation machines were targetted to the labor-intensive task of generating artillery tables, these devices were called "electronic computers."

    Every time I see a flame session about the lack of women in computer engineering, I find it ironic that the word "computer" itself is an artifact of a group of geek girls.

  • Alan Turing worked on speech encipherment at Bell Labs for a couple of months in 1943. It seems that he didn't actually work on the SIGSALY system, but did a technical appraisal of it for the British government.

    Turing came up with Delilah after he returned to the UK. It was a much smaller and simpler device, but it was never put to any practical use.

    More info can be found in Andrew Hodges' book "Alan Turing: the Enigma" [turing.org.uk]. (Hodges maintains The Alan Turing Home Page [turing.org.uk], referred to in the parent comment.)
    --

  • If you're going to have to produce and distribute "noise key" records in the first place, why not simply combine the two signals (the voice and the noise) in an analog fashion before transmission, and then do the same process in reverse at the other end? This would have been MUCH simpler (meaning they could have deployed it sooner).

    Because the signal that is transmitted will be subjected to additional noise. This has a much greater effect on the encrypted signal, if the encryption is of any reliable level at all, than it does on plaintext sound. The digitization had the same advantage then it does in modern systems, noise can be completely removed which is a necessary first step toward reconstructing the plaintext.

  • I guess it's okay that there's now more speakers of Klingon than of Navajo. We can still have secure communications without our warm, whirring little friends.
    --
  • Or ~55 years for those of us who aren't stuck in the 80's...
  • Oops. I meant More speakers of Klingon than Navajo. [theonion.com]
    --
  • Was this system discussed in Cryptonomicon or one of the other dozen /. reads I've been through this year?
  • damn Germans! anyway makes you wonder what kind of current crypto techniques are in use...blows my mind...we probably can't imagine....

  • Meanwhile, in the 21stC (Score:3)

    by BenBenBen ( 249969 ) on Friday January 12, 2001 @09:00PM (#510293)

    If you want to look at cool modern-day intel. stuff, the GCHQ website [gchq.gov.uk] is actually pretty detailed.

    The largest LAN in Europe, one of the highest data storage capacities in the world, and free healthcare =)

    Also, they seem to think they can pass secret instructions to spies the world over by inserting 'random' bold tags on one of their pages [gchq.gov.uk]

    Ben^3

  • Navajo Code Talkers in World War II (Score:1)

    by Anonymous Coward
    Here is a link for you: Navajo Code Talkers in World War II [navy.mil]
  • Keep looking in old articles and you will find that patents in telephone encryption often have to cite the work of the recently deceased movie star Hedy Lamar as prior art. She worked with engineers in her husband's company to develop telephone encryption. Her ideas were innovative and sound, but barely workable in WWII technology. The same could be said of the system described in this /. article.
  • I would like to first point out that I do not know any and certainly do not speak Klingon.

    Now, as Klingon was likely contrived by English speakers I would imagine it would have a fairly simple to decode structure, idioms (if there even are any), and all verbs and nouns would probably have 1:1 equivalents, with the exception of stupid star trek shit.

    As such, I doubt Klingon speakers would make very good code talkers. Although putting them in the line of fire is not necessarily an entirely unattractive proposition.

    Now once again, to reiterate, I may be wrong, as I do not know any Klingon, nor do I ever intend to.
    ---
    "You just stranded one of the world's greatest leaders in San Dimas!"

  • Analog scrambling systems have a long history of being easily broken. It is just too easy to recognize and extract the distinctive patterns of human speech.

    A vocoder converts an analog voice signal to a digital bit stream, which can easily be encrypted by any number of devices. The advantage of a vocoder, in comparison to the PCM encoding techniques used in the telephone system, is a much lower bit rate. A vocoder constructs a model of the human vocal tract, and transmits the parameters of the model at a periodic rate. The receiver uses these parameters to construct an electronic replica of the speaker's vocal tract, which produces the audio that the other party hears.

  • incredible (Score:5)

    by burris ( 122191 ) on Friday January 12, 2001 @10:20PM (#510298)
    The system required that the key be used in twenty millisecond segments. Therefore, it was necessary for each record to be kept in synchronism within a few milliseconds for fairly long periods of time (one hour or so). This was accomplished by the use of very precisely driven turntables. The turntables themselves were remarkable machines. Each was driven by a large (about thirty-pound) synchronous electric motor with hundreds of poles. The motor was kept in constant operation, and the power for it was derived directly from dividing down the terminal's frequency standard. The frequency standard was a 100 kHz crystal oscillator. The accuracy of the standard had to be maintained within about one part in ten million so that the system would stay in synchronism for long periods of time. The system frequency standard could be corrected by comparing it to an available national frequency standard (which was WWV in the U.S.).
    For reference, the clock in a high performance modern digital audio device is only accurate to within about 1-10ppm. A cheap consumer device like a CD player or walkman type DAT deck will only have about 50-100ppm accuracy. This is 0.1ppm, in 1943, for a freakin' record player. Absolutely amazing.

    Burris

Slashdot Top Deals

"Why can't we ever attempt to solve a problem in this country without having a 'War' on it?" -- Rich Thomson, talk.politics.misc

Close