December 2003 - cypherpunks-legacy

Native American Encryption?!
by Arthur Chandler 17 Dec '03

17 Dec '03

I remember hearing many years ago that one of the branches of the armed aervices during WWII used a native American -- Cherokee? Apache? -- language to communicate. The Japanese (the story goes) went nuts trying to crack the code -- unsuccessfully. Two questions: 1) Can anyone give me a reference for this story (assuming it's true)? 2) If one used a natural language for encryption, and the would-be code crackers did not know it was natural language (say, Hittite), could they crack it? I seem to remember that hieroglyphs were undecipherable until the Rosetta Stone was discovered. But maybe current techniques would do a better job....?

2 1

Secret network
by Matthew J Ghio 17 Dec '03

17 Dec '03

There is something else that we can use for secret communications and that is ... (don't laugh, I'm serious) - water pipes. Many places still use metal pipes, thru which radio waves can travel quite well. I read an article in a science magazine a few weeks ago, where a team of researchers were tracking various electromagnetic emissions in residential households. They fould that a lot of background magnetic fields and radio emissions were coming from the water pipes. When the neighbors used certain appliances, they could detect it because the EM emmissions were picked up and carried by the metal water pipes. And that's was just background noise. Imagine what we could transmit by intentionally broadcasting signals thru the metal pipes! Also, most of the "noise" was low frequency, and since our transmitters would be high frequency, it would probably work pretty well (unless, of course, you had plastic pipes). Most of the local net traffic could be handled thru the pipe-network, while the radio transmitters, lasers & microwave stuff could be used for longer distance stuff.

1 0

Spread Spectrum Decoys
by Hastings＠courier8.aero.org 17 Dec '03

17 Dec '03

Part of the "hunt and destroy vs. cost to deploy" equation thread about spread spectrum networks was forwarded to me. How about putting several cheap noise generators on the air as decoys? This might tip the cost balance in favor of the network. The Julieboard Direct Digital Synthesizer board, which only needs numbers fed to it from a PC printer port or microcontroller output, can be bought wired and tested for $125. Production in volume should be cheaper. So who is hunting down transmitters and killing them, and risking a confrontation with an armed opponent, for this paltry sum? Kent - <hastings(a)courier8.aero.org>

2 1

TEST: ignore.me
by Stanton McCandlish 17 Dec '03

17 Dec '03

testing blahb lahblahblabhblhalblah blah sorry folks... -- -=> mech(a)eff.org <=- Stanton McCandlish Electronic Frontier Foundation Online Activist & SysOp NitV-DC BBS 202-232-2715, Fido 1:109/? IndraNet 369:111/1, 14.4V32b 16.8ZyX

1 0

Re: Spread Spectrum Decoys
by swc＠uc1.ucsu.edu 17 Dec '03

17 Dec '03

Building boards from scratch always sounds a lot cheaper than it is. You'll find it hard to build any kind of electronics for less than $126 5. The manufacturing cost almost always exceeds the component cost, and you always needs l lots of little components that you didn't think of at first (miscellaneous glue logic, power supply, enclosure, etc.). ------------------------------------------------------------------------ Stuart W. Card, Consultant, Card & Associates -- Research & Development Box 153 RR 1 Newport Rd Utica NY 13502 315-735-1717 / FAX -8469 swc(a)uc1.ucsu.edu or cards(a)top.cis.syr.edu "Who is John Galt?"

1 0

Re: Breaking DES
by doug＠netcom.com 17 Dec '03

17 Dec '03

> pmetzger(a)lehman.com said: >(2^56)*8 = 576,460,752,303,423,488 > >(2^56)*8*1000/(10^9) = 576,460,752,303 I was originally assuming a one byte result per calculation and gave a hashing justification. (I glossed over the overhead for this approach, but we could analyze this, too, if anyone's interested.) You however are multiplying by 8 for no clear reason. Feel free to explain, but the way I see it, the calculation is: (2^56 calculations) * (1 byte per) * ($1000/disk) / (10^9 bytes/disk) ...which comes to 72 billion dollars. This is significantly larger than my first calculation, and the difference is purely due to roundoff error, because I said that 2^56 = 10^16, where it's actually 10^16.8576. 10^16 * $1000 / 10^9 does indeed equal my original figure of $10 billion, so it's just that I should have left another digit of precision in rounding the exponent. Now I agree that 72 billion dollars is a lot. Even 10 billion is a lot. But all I was trying to establish was that these numbers are not *completely* impossible, because you were giving Karl a hard time about this. Ridiculously expensive is a very different thing than *impossible*. There are schemes that result in figures like 10^9 billion dollars... *that* I call plainly impossible. I also projected forward 10 years to let prices come down by a factor of 10...that's another way of underscoring *possibility*. Over the last 10 years disk drive prices have fallen significantly more than a factor of 10 per megabyte, so I am being quite cautious here. A factor of 30 is justifiable, but I won't go that far, I'll continue to be conservative. Even your own figure of $576 billion becomes $57.6 billion in 10 years, which is merely too expensive, not *impossible*...if WW III were underway and such a project were of critical importance, it would happen... $50 billion would not be too much under *those* circumstances. That's the difference between "impossible" and "expensive". >Lets think for a moment, shall we? You are encrypting every possible >block with DES, which results in lots of random blocks. You really >want to search through the lot of them serially without any indexing >whatsoever? Seems like you haven't thought this out. If you want to critique this part of my back-of-the-envelope, its weakest part is the sorting, in which it is very hard to effectively serialize disk access. For the benefit of the doubt, let's give that a factor of 100 slowdown... so that 10 years from now we have an *average* disk transfer rate of 10^7 bytes per second for this algorithm rather than the 10^9 that I was assuming. I think it could be done faster, but even so, this increases the time from 1 hour to 4 days...still not an impossibility, just not as *nice* as 1 hour. Again, I need only establish possibility for the algorithm Karl related; I'm not saying the NSA *will* do this. >Oh, now I just realized -- you are going to have to store each source >block with each output block. That means that even if you don't do any >indexing, you are going to need twice the disk space I just mentioned, >or over $1 TRILLION in disk for a very slow DES cracker. Ok, so figures are doubled...my 72 billion becomes 144 billion. Pretty expensive. Not *impossible*. >Sorry, but you lose. I never claimed this was likely...all I was after was to see whether it worked out to e.g. 10^9 trillion dollars...*that* I would call impossible. >Never designed a disk system, have you? Actually, yes I have; I've been a hardware and software systems architect in all kinds of different subspecialties. The fact that we may disagree doesn't make me an idiot...hell, I may even make drastic mistakes and say things that *are* idiotic. It still doesn't make me an idiot...it would make me "someone who made a mistake". Considering that I was doing a quickie back-of-the-envelope, I'm not even embarassed about such mistakes. No one else did an estimate. 2^56 *sounds* ridiculously huge; I'm content to be within a factor of 100 in showing that it is merely quite expensive. I daresay that you yourself have a better feel for the expense now than you did when you first critiqued Karl's post. Flaming me is a poor way to win an inherently technical argument. Stick to the point. >Sorry, but you can't actually >read a disk as fast as you can read RAM. Caching only works if you >have frequently accessed blocks -- if you are reading a whole disk you >can't go faster than the disk transfer rate no matter how many gods >you pray to. Your technical credibility is rapidly plunging. It is true that you can't do better than the average transfer rate, and here you have a valid point, I neglected this. It would be unrealistic given the other assumptions to assume better than 10^7 bytes per second transfer rate with technology 10 years hence. In fact even 100 megabytes per second might seem high to you, so let's call it 50Mb/s (surely a very conservative figure), for a total of 20 times slower than I estimated. That increases the 4 days to 80 days. Not very nice...but *possible*. You see the pattern here...you are raising valid technical objections, a whole series of good points that I glossed over with my back-of-the- envelope calculations. But even so, it doesn't change my basic point that the approach is *possible*. You need to find a factor of perhaps 1000 in cost and a factor of perhaps 1000 in time in order to demonstrate that this approach is inherently *impossible*. The fact that you spot flaws in my back-of-the-envelope also doesn't mean that it's called for to flame me. Again, let's stick to technical discussion. >Guess you can't read my calculations, can you? Tsk, another flame. >> Skip the sarcasm and pick a different quantity discount. If you don't >> like my 90% discount for quantity 1 million disk drives, pick another >> one. > >Sort of like saying "if you don't like the laws of physics, pick >different laws"? Even for a flame, I don't get this. I said, if you think that a 90% discount for quantity-million is unrealistic, tell me what discount you think *is* realistic. That's a valid question. The $1000 per gigabyte drive is roughly accurate *today* in quantity *one*. The higher the quantity you buy, the better a discount you get; that's the way it works, and I'm sure you know that as well as I. So perhaps my 90% discount is overly optimistic...fine, I say...tell me a different figure. If you say 10% I'll argue. Anything between 10% and 90% is conceivable, so pick your figure. It still doesn't affect the bottom line argument as to whether the algorithm Karl mentioned will be possible in 10 years. It clearly would be very very expensive. It would also clearly *not* be completely impossible. Karl posted something which is theoretically reasonable but that is nontrivally expensive even ten years from now. He deserves credit for discussing a theoretical possibility which is even marginally conceivable. He does not deserve a harsh response...and I think you *were* harsh to him. Do me a favor and skip the flames in your future responses; they're not very much fun. Doug -- Doug Merritt doug(a)netcom.com Professional Wild-eyed Visionary Member, Crusaders for a Better Tomorrow (The above is a joke; the following are mailing lists:) Unicode Novis Cypherpunks Gutenberg Wavelets Conlang Logli Alife HC_III Computational linguistics Fundamental physics Cogsci SF GA VR CASE TLAs

1 0

Re: Spread-spectrum net (vulnerability of)
by swc＠uc1.ucsu.edu 17 Dec '03

17 Dec '03

Has anyone pointed out that the ideal location for these packet repeaters is in the cars of as many people as possible? ------------------------------------------------------------------------ Stuart W. Card, Consultant, Card & Associates -- Research & Development Box 153 RR 1 Newport Rd Utica NY 13502 315-735-1717 / FAX -8469 swc(a)uc1.ucsu.edu or cards(a)top.cis.syr.edu "Who is John Galt?"

2 1

Re: Native American Encryption?!
by doug＠netcom.com 17 Dec '03

17 Dec '03

tcmay(a)netcom.com (Timothy C. May) said: >They were the Navajo (or Navaho) code talkers. Kahn's "The >Codebreakers" has a discussion of this, as I recall. Probably the >"Encyclopedia Britannica" will also a mention of it. It's mentioned in a bunch of crypto sources, but coincidentally, PBS just ran an entire show about this precise subject a couple weeks back. Not just passing mention...the whole show was about this. Therefore talking to PBS people would likely yield some good info... and possibly videos. >> 2) If one used a natural language for encryption, and the would-be code >> crackers did not know it was natural language (say, Hittite), could they crack >> it? I seem to remember that hieroglyphs were undecipherable until the >> Rosetta Stone was discovered. But maybe current techniques would do a >> better job....? > >These are codes, not ciphers, and are of course not very secure. Err...."of course"??? Codes are (all else being equal) quite a bit more secure than ciphers. (or do I have "code" and "cypher" reversed...whatever. :-) There's a classic SF story, whose title I forget, about anthropologists trying to figure out writings of a dead species on Mars. The table of elements finally proved to be the Rosetta Stone equivalent. The interesting thing about codes, which in a sense includes all natural languages, versus ciphers, is that code systems represent semantics. If the underlying semantics is radically different than what the code-breaker knows...too bad. >Germans and Japanese in WW II obviously did not have enough time to >find native Navajo speakers, and I suspect few books on that language >were available at that time, hence the scheme was temporarily secure. > >Otherwise, forget it. It was more complicated than that. They used a hybrid system that native Navajo speakers could not decrypt, because the system used not only Navajo, but on top of that, arbitrary (and newly invented) metaphors for concepts, and (newly invented) puns to represent ciphers, too. It is certainly true that part of the security was "through obscurity", but (A) that part was effective..."security through obscurity" can be effective over short periods of time...and (B) they layered ciphers on top of codes. The obscure linguistic aspects of Navajo vs. other modern languages is also said to have played a part, but I haven't researched this yet, so I won't comment. Arthur Chandler <arthurc(a)crl.com> said: >I seem to remember that hieroglyphs were undecipherable until the >Rosetta Stone was discovered. But maybe current techniques would do a >better job....? In the absence of a Rosetta-Stone-sort-of-thing, we're still lost. For instance, the Easter Island hieroglyphs are still completely mystifying. Some of Nyquist's mathematical results are still classified, so one never knows, but... Arbitrary semantic systems encoded in writings are not decipherable, period, barring some breakthrough in mathematical semantics...don't hold your breath. :-) Doug

1 0

Re: Spread-spectrum net (vulnerability of)
by doug＠netcom.com 17 Dec '03

17 Dec '03

jkreznar(a)ininx.com (John E. Kreznar) said: >Not even the antenna should be visible. Remember, for spread spectrum radio >detectable only by the intended receivers (which use the correct spread- >spectrum code for coherent detection), the power density should be hardly >distinguishable from the ambient noise. This is true only for casual observers. If the FCC were after you, they could most definitely triangulate on unusual noise sources as well as they could a coherent signal. Power signals are a giveaway. >One way to arrange this is to put the antenna inside of a physically >secure perimeter, outside of which the power density is too low for >noncoherent detection. The perimeter can be optically opaque (e.g. a >building), as long as it leaks enough r.f. in the direction of the >receiver(s) for coherent detection. Preventing noncoherent detection >may often require _attenuation_ of an otherwise too-powerful signal, >and the building may serve part of this function. Good enough as far as it goes. But this implies a large drop in efficiency of the transmitted signal. That's not a stopper...*if* you've got power to spare. But that implies enough power for bad guys to triangulate your noise source...ouch. If they pin you down to within a building, you've lost. There are other approaches...phase-sweeping...phase-conjugation... Doug

1 0

Re: Breaking DES
by doug＠netcom.com 17 Dec '03

17 Dec '03

Thank you, Perry, for some good comments that were flame-free. I personally appreciate that, especially considering that your comments are both apropos and good food for thought. Doug >pmetzger(a)lehman.com said: >Karl Lui Barrus says: >> So maybe it's only of theoretical interest, sort of like differential >> cryptanalysis against the DES - which requires 10^47 chosen >> plaintexts. >> >> Why don't you mail Biham and Shamir that their method sucks. It's >> fairly infeasible as well. > >It *IS* infeasable, and they realize it. The breakthrough was >differential cryptanalysis itself, and the discovery that DES was >fairly resistant to it. The fact that they made ANY crack in it was >kind of neat, by the way. > >A huge number of chosen plaintexts is of course pretty much not >possible in practice, especially since you might not get any chosen >plaintexts at all!

1 0