Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!seismo!caip!meccts!mecc!sewilco From: sewilco@mecc.UUCP (Scot E. Wilcoxon) Newsgroups: net.crypt Subject: Re: Naieve Inquiry: Streaming Cyphers ( Message-ID: <540@mecc.UUCP> Date: Sat, 2-Aug-86 00:02:19 EDT Article-I.D.: mecc.540 Posted: Sat Aug 2 00:02:19 1986 Date-Received: Sat, 2-Aug-86 22:14:59 EDT References: <1065@ogcvax.UUCP> <25900002@okstate.UUCP> <189@inuxa.UUCP> Reply-To: sewilco@mecc.UUCP (Scot E. Wilcoxon) Organization: MN Ed Comp Corp, St Paul, MN Lines: 56 In article <189@inuxa.UUCP> michel@inuxa.UUCP (Alan Michel) writes: > > >How about taking parts of the key, and generate two initial seed values for >long psedo-random sequence generators in a pre-determined way. >Then use bits from one sequence to indicate >when to place a random bit into the data stream and >the other random sequence (or even later bits in the same sequence) >to indicate whether that random bit should be a 0 >or a 1. ...[example omitted] >WARNING - I know next to nothing about real world cryptography. >This seems like an obvious and quick way to pad data, but I do not know >how easy it is to crack, or how much it may really increase/decrease >security. Just a thought. I'm an amateur myself, having only designed and cracked a few dozen simple digital codes. His proposal sounds good for a distraction within or around a better code, but I can see two weaknesses: The first bits are less random than the rest are, and the more random (encrypted) the 'data stream' is, the better it will work. The first weakness is visible with ASCII English text as the input. Take the first bytes and try omitting the first bits to see if a printable character shows up. Then look at the next few bits to see what needs to be done to get the most likely English character. Keep that up and look for patterns in the identified encryption for clues to the keys (easier if the pseudorandom algorithm is known). The example in the original article used only 1 of 8 bits random, so good guesses will quickly be confirmed by real words showing up. Many more random bits are needed to obscure the data. When English text is encoded, the left-side weakness is obvious. But if the input data stream is pseudorandom (ie, encyphered data) then this blending of pseudorandom data is a bit more distracting. The question is whether or not analysis can identify the two different patterns, just as there are easy ways to identify the good guesses with English data. Perhaps analysis would show two different kinds of pseudorandomness which then could be separated. Or as someone recently pointed out perhaps sync characters or other artifacts of the data stream (I've easily broken several unknown algorithms on systems which used a certain line terminator or space padded data). Or there's always the possibility that a weakness in the key-to-sequence algorithm can be exploited to find either the key or part of the sequence. (Should we amateurs ramble in here?) While I have the podium, since my last "coding by location of typeset characters" article I've wanted to observe that this newsgroup is unusual as users of discussed methods tend to not identify themselves for obvious reasons. It's intriguing to know that ideas which generate no response may be idiotic or everyone who would comment won't because they've been using the method for years :-) -- Scot E. Wilcoxon Minn Ed Comp Corp {quest,dicome,meccts}!mecc!sewilco 45 03 N 93 08 W (612)481-3507 {{caip!meccts},ihnp4,philabs}!mecc!sewilco NASA:"Earth uninhabitable in 500 years." Welcome to tropical Minnesota.