December 2003 - cypherpunks-legacy

for interesting people -- Sunday NYTimes Mag re: encryption
by Lee Tien 17 Dec '03

17 Dec '03

Sunday's NYTMagazine had an amusing bit in its "Sunday" column, titled: HTIJX KTW PNIX "The scrambling and decoding of electronic communication is one of the hottest issues in cyberpolitics. The Government would like to have the key to all communications. Export of cryptographic software is restricted. Both software companies and independent groups like the Cypherpunks argue for the rights of individual privacy." "All of which makes Microsoft's new Creative Writer software for children intriguing. It includes some primitive cryptographic tools to enable kids to encode what they write ... to hide it from curious siblings and F.B.I. agents." What surprised me was the next line -- "Theoretically, this product would come under export restrictions, even though the code would not challenge a spy, let alone a bright 9-year-old." After explaining what a Caesar substitution cipher is, the bit goes on to note that "Presumably, Microsoft will not face any difficulties in exporting its product even though it comes from the K.G.B. (Kids Games Business) at Microsoft. On the other hand, it does instruct the youth of America in the basics of electronic privacy." In case you're wondering, the encrypted phrase is "Codes for Kids." Lee Tien

1 0

Crypto-dongle redux
by ccat＠netcom.com 17 Dec '03

17 Dec '03

I think the idea of a standard user configurable crypto-engine is a good one..that way people could choose (and upgrade) algorythms.. sort of like modern DSP based highspeed modems..(It could also have a socket for ROM based firmware cartridges..(RSA,IDEA,etc..) and a magstripe reader and keyboard for key entry..(This would make trading keys easy..) I think this discussion was had before,about a year ago.. Would anyone care to elaborate on it? -Chris.

1 0

Comments on violence
by Carl Ellison 17 Dec '03

17 Dec '03

I don't know if legalization would prevent the violence at this point. It would probably reduce street crime but I'm concerned that our population is getting adicted to the thrill of violence (movies, TV news, video games, ...) and it might take a kind of drug treatment program for the whole nation to do anything about it. My second, bigger concern is that the government is using this violence as an excuse to attempt to remove our rights. The one I'm most concerned about is the right to privacy in communications through cryptography (since I'm a computer system architect and cryptologist). We have always had the right to invent and use our own strong cryptography (for the whole 4000 year history of cryptography) and yet the administration (no doubt prompted by the NSA) is now trying to claim a right to read all private messages [cf., the "Clipper" proposal], citing violence (drug dealers, snuff movies, etc.) as the need, without establishing a link -- just using the reference as a scare tactic to get voter adrenalin flowing so that they'll feel pro-law-enforcement. This is a major conflict between government and civil rights, coming to a head as we speak. A spotlight on this attempt by you would be appreciated. Thank you. Carl Ellison Sr. Technical Consultant Stratus Computer Corp. / ISIS Distributed Systems

1 0

Re: Randomness of a bit string
by rjc＠gnu.ai.mit.edu 17 Dec '03

17 Dec '03

Tim writes: >But can he ever say "I can prove the number is random"? No. There's >always some chance an even-cleverer puzzle solver will find the >pattern, the key that unlocks the randomness. For example, most >ciphertexts pass nearly all statistical tests for randomness, "look" >random, and even _act_ like random numbers (recall the Blum-Blum-Shub >pseudorandom number generator and how good it is). But simple >application of the key turns the seemingly random >"100010001010110010101" into "ATTACK." But can we say that "100010001010110010101" has been ``compressed'' into "ATTACK"? How do we know? Let IC(x) stand for the amount of information storage used by x. Is IC(100010001010110010101) > IC(ATTACK) + IC(key) + IC(algorithm)? It is not at all clear that this relationship would hold. (in fact, I don't think it will even begin to work out unless the cyphertext is much longer than the plaintext) So in fact, cryptorandom numbers can be considered incompressible if you take into account the algorithm required to perform the operation -- just as if I had used a 100 terabyte dictionary to compress via lookup, or better yet, a one time pad. -Ray All of this is meaningless anyway. Information theory was proven wrong by WEB technologies when they invented a compression program that can recursively compress any input data down to 64k. Harddrives are now obsolete. -- Ray Cromwell | Engineering is the implementation of science; -- -- rjc(a)gnu.ai.mit.edu | politics is the implementation of faith. --

4 4

Re: NSA museum now open, if you can find it
by Dave Hart 17 Dec '03

17 Dec '03

---------- | From: Hal <netmail!hfinney(a)shell.portal.com> | | That museum sounds fascinating. I got to visit the NSA's so-called | "Friendship Annex" once on business. This is not at Fort Meade itself, | but a few miles away, to keep the impure and unclean away from the holy | temple itself. | | Whoever named this place had quite a sense of irony; [...] As documented in _The Puzzle Palace_, the name derives from Friendship International Airport, now known as Baltimore-Washington International Airport. The Annex buildings are at the end of one of the runways. Dave Hart davehart(a)microsoft.com Not a Microsoft spokesperson

1 0

Re: Could I possibly be deleted, please...
by Al Billings 17 Dec '03

17 Dec '03

On Mon, 24 Jan 1994, Lance Ware {VOXEL Guru} wrote: > Please remove me as well . . . Try using the admin address. The list can't remove you. We only see your messages. -- (*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*) | Al Billings aka Grendel Grettisson | "You are, each one, a priest, | | mimir(a)io.com | Just for yourself." | | Sysop of The Sacred Grove (206)322-5450 | | | Admin for Troth, The Asatru E-Mail List | -Noble Drew Ali- | (*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)(*)

1 0

quote of the day
by Carl_Ellison＠vos.stratus.com 17 Dec '03

17 Dec '03

Only those means of security are good, are certain, are lasting, that depend on yourself and your own vigor. - Machiavelli

1 0

Provability and Randomness
by collins＠newton.apple.com 17 Dec '03

17 Dec '03

Entropy is relative. A string is `random' (with respect to an observer) when the probability of correctly predicting the next symbol of the string is arbitrarily low (e.g., size_of_the_alphabet^-1). Entropy, and therefore `randomness' can only be considered in the presence of symbol probabilities... and therefore prejudicial knowledge i.e., a context (algorithms, models, history, whatever). Different contexts --> different probabilities --> different quality of randomness. * Absent a context, there is no such thing as `randomness'. Posit two identical contexts, sender and receiver. Sender transmits a `random' string to receiver. (Beeeeeeeeeeep. Sorry, that was the warning that sounds whenever I fib). The sender can only send a random string if the reciever doesn't already `know' that string or doesn't know which string the sender will transmit. If the sender knows something that the reciever doesn't then the contexts are not identical. * Absent differing contexts, there is no such thing as randomness. A fair coin toss can be random because you_before_the_toss and you_after_the_toss are different contexts (reciever and sender, respectively; one of whom knows the outcome). Posit two disjoint contexts, A and B. A transmits a message to B, who has no information in common with A. B has no context with which to predict the first symbol that will appear and thus it is always random. As symbols appear, B builds a model of A... and thus acquires knowledge of A (i.e., a shared context). By the end of the message, B might be predicting quite well. If B can't build any model of A's behavior at all, then B will share no context with A; won't be able to predict characters; the string will remain random. * Absent shared knowledge (overlapping context), all information is random. Imagine that B's shared knowledge with A takes the form of a program to output a prediction of the next symbol A will transmit. This program---however large it is and however it might work inside---is nothing more than B's model of A. When B has no knowledge of A, this program is essentially `empty'. It contains no information, and can make no predictions better some arbitrary limit (e.g., size_of_the_alphabet^-1). The program learns from each symbol transmitted by A, thus a good (and portable) measure of the `size' of the program is how many symbols it has seen. Let us say that this program sees every symbol A ever transmits to B (numbered from 1..n), and thus during it's life it will actually be n+1 different programs (numbered B0..Bn of size 0..n, respectively). Imagine that you can ask any one of these programs to predict any symbol from A. Thus you could ask B3 to predict symbol 4 (exemplary of the normal case) or you could ask B5 to predict symbol 1 (which it could, of course, do perfectly, having already seen symbol 1). Now we have a new definition of randomness. A string is random with respect to B if no program of B shorter than the string can predict it with success greater than our arbitrary threshhold (which is typically defined by the performance of B0). If A is sending a passage from a well known book, and B `discovers' this after receiving symbol 20 and can access the text of that book, B20 suddenly becomes a very good predictor of many future symbols. The string is not random. But it _was_ random to B0, and B1 and perhaps less for each successive symbol. B20 is a different context than B0. It has different knowledge, different probabilities and therefore perceives a different quality of randomness in A's message. B20 is still only a program of `size' 20 (i.e., you don't count the size of the book in B). This is easily demonstrated if you imagine what happens when A sends a message that is a deterministic algorithm for producing a an infinite stream of symbols, followed by the stream it generates. If this algorithm requires i symbols to express, then Bi is a perfect predictor for all subsequent symbols. Bi is clearly of size i (there is no external book for us to add to the size of Bi). In fact, no matter what message A sends, B considers it an algorithm for generating predictions of future symbols. Thus A is actually sending B a sequence of programs (each a prefix of the next, and thus not re-transmitted) B1, B2, ... Bn (but remember, these programs execute in the context of B's knowledge... thus their predictions are not `universal'). This just brings our notion of programs, program length and prediction around to the other side and lets us summarize: * A string is random with respect to B if the string itself is the shortest program with which B can generate that string. ... or qualitatively * The randomness of a string Bn with respect to B is an inverse of the quality of the predictions B can make of Bn from the strings B0...Bn-1. We rely on the `relativity' of entropy. Codes and cyphers can't function without it. The difference between your context and that of an attacker (you know the key or codebook) is what makes the message meaningful only to you (hopefully it will still have _some_ information you couldn't guess before reading it). Randomness is relative, thus there is no universal randomness measure for a string, thus there can be no proof that a string is universally random. You can easily measure the exact entropy of a string with respect to a very formally defined context (one where you can produce exact predictions). This is useful, but reveals nothing about the quality of the predictions a different, even similar, context might make (Just one symbol is the difference between B19 and B20 above; the string was random to one but not the other), It reveals nothing about models we can't describe so perfectly (like human thought). * There is no algorithm for deciding if a string is universally random. In a less obvious leap, it is only by comparing the predictions of Bi with Bk that a string of length j (i < j <= k) can be shown to be random with respect to Bi. Thus: * There is no algorithm shorter than the string itself for determining if a string is random with respect to a given context. Not exactly Q.E.D. but close enough for rock `n roll. Scott Collins | "Few people realize what tremendous power there | is in one of these things." -- Willy Wonka ......................|................................................ BUSINESS. voice:408.862.0540 fax:974.6094 collins(a)newton.apple.com Apple Computer, Inc. 5 Infinite Loop, MS 305-2B Cupertino, CA 95014 ....................................................................... PERSONAL. voice/fax:408.257.1746 1024:669687 catalyst(a)netcom.com

1 0

The Packwood Memorial Diary Server
by jim＠bilbo.suite.com 17 Dec '03

17 Dec '03

I often see threads debating whether the courts have the right to compel you to reveal your encryption key. Some say yes, some say no. Under the assumption that courts *do* have the right (or power) to force you to reveal your key, the problem now becomes: How can you prevent the government from proving you have encrypted documents? One Answer: Don't keep encrypted documents in your possession. Somebody could create a Document Server to store encrypted documents. Users would somehow get an anonymous Document Server account number when they subscribe to the service. Users would use the remailer system to send encrypted documents to the Document Server. The account numbers would be used to organize the document database and for billing (the tricky part). Given a Document Server, the problem now becomes: How can you prevent the government from proving you use a Document Server. This seems like a strictly technical problem, unlike the "can they compel you to reveal your key" problem. Jim_Miller(a)suite.com

3 2

Re: REMAIL: Cover traffic
by jim＠bilbo.suite.com 17 Dec '03

17 Dec '03

Arsen Ray Arachelian writes: > ...Even better, have each remailer send a specific > number of messages to each of the other remailers on the > network. These messages would be bogus messages, > however, there would be a fixed number of them. > > If a real nym message arrives, it is sent to the next mailer > up the chain, as part of the n (n-1 now) that are bogus. That > way a spook couldn't tell where a message was going since > he couldn't count the number of messages going out of the > mailer. > > Also if a target remailer has n real messages to be sent to, > any messages over that assigned packet size of messages > get spooled for the next round of bogus mail. This way each > remailer will send exactly n messages to every other > remailer on the net every specified period of time. > I like this idea. It seems to use fewer CPU resources than having a remailer route a bogus message through a random set of other remailers and back to itself. If I understand the encrypted remailer scheme correctly, the "route through random set" mechanism requires a remailer to enclose a bogus message in a set of nested digital envelopes (one for each remailer in the random remailer set). The "round-robin send to peers" mechanism only requires the remailer to create one envelope per bogus message. I also like the idea because it seems easier to analyse, and therefore easier to describe/formalize the properties of the system as a whole. Jim_Miller(a)suite.com

1 0