home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.crypt
- Path: sparky!uunet!newsgate.watson.ibm.com!yktnews!admin!wo0z!lwloen
- From: lwloen@rchland.vnet.ibm.com (Larry Loen)
- Subject: Missing subject header
- Sender: news@rchland.ibm.com
- Message-ID: <1992Aug17.222410.10011@rchland.ibm.com>
- Date: Mon, 17 Aug 1992 22:24:10 GMT
- Reply-To: lwloen@vnet.ibm.com
- Disclaimer: This posting represents the poster's views, not necessarily those of IBM
- Nntp-Posting-Host: wo0z.rchland.ibm.com
- Organization: IBM Rochester
- Lines: 50
-
- In article <1992Aug12.185029.13607@mintaka.lcs.mit.edu> Eric Knight writes:
-
- >Yes, I was attempting to create an encryption that would become
- >I/O bound as the limit to its speed. Most other systems become
- >CPU bound (DES, great example.) I/O is significantly slower than
- >CPU and that causes problems repeatitively trying to guess it.
-
- Well, we're all for a fast system. Speed is, unfortuately, one of the later design
- criteria. The system has to take a LOT longer to solve than it does to encrypt.
- If breaking it takes only 100,000 times as long to do as encrypting it, then it
- will be a failure in virtually any application I can imagine. If it doesn't work,
- fast run time will not be impressive.
-
- Cryptography has great appeal to any number of people.
- It looks so easy! And, indeed. Nothing is easier than proposing a new
- encryption algorithm. Nothing is harder than figuring out whether it is any
- good.
-
- I am hardly the top gun in the universe on this topic, but I have learned
- a considerable respect for the need to study it. It is very highly probable
- that unless you get out and study a bit, you will produce a system that
- runs fast, scrambles the data impressively to the eye, with cipher output
- output passing all kinds of statistical tests for randomness, has a 500 gabillion
- bit keyspace, but can be solved by someone you've never heard of that knows what
- they are doing. Oh, yes. The system did stump me and all your friends, too. Just
- didn't stump the guy you really needed to fool. The history of cryptography
- simply abounds with examples of this. Read "The Codebreakers" by David Kahn
- (in most libraries and quite good).
-
- I also highly recommend that you acquire some experience in breaking cryptography
- systems. There are no scientific proofs for what makes a system good. Lots of
- mathematical things hold "in general", but fall apart because some niggling little
- special case comes up that an attacker can expect to exploit sooner or later,
- and that brings the whole edifice down. There's even a semi-formal name for
- this. It's called "entering" the cipher and once you have done so, a great
- percentage of systems slowly but surely break as a whole. Or, enough to
- steal the PIN on someone's ATM card or the like.
-
- Unless you have broken quite a few of the nontrivial, classical cryptography
- systems, your chances of developing something good are pretty small. There's
- so many things you have to guard against. And, only practical experience and
- lots of study will give you a clue as to what they are. Do you know whether
- your system is vulnerable to various forms of advanced mathematics? Do you
- know, given that a lot of data is boilerplate and may be correctly guessed,
- how many bits of data that have to be guessed to defeat you? It's a long list,
- and I've only typed in a little bit of it. . .
-
- --
- Larry W. Loen | My Opinions are decidedly my own, so please
- | do not attribute them to my employer
-