Re: WEAK3 -- A Layman's Data Encryption Algorithm

Mok-Kong Shen (mok-kong.shen@stud.uni-muenchen.de)
Mon, 20 Jul 1998 12:08:57 +0100

• Next message: William H. Geiger III: "Re: MD5 partial collisions"
• Previous message: Mok-Kong Shen: "Re: DES cracker built"
• In reply to: Bruce Schneier: "DES cracker built"

Ian Clysdale wrote:
>
> > Owing to the non-trivial programming efforts required, WEAK3 does not
> > appear to well satisfy the requirements of a poor man's environment,
> > though it may be quite useful in normal environments, being
> > independently implementable from scratch in one to two weeks with
> > average programming experience.
> >
> I don't find that particularly impressive. I think that taking two weeks to
> implement ANY of the common block algorithms from the specification is close
> to excessive, for anyone with a reasonable degree of programming experience.
> Now, you can spend a huge amount of time in optimization, but the basic
> algorithms are usually rather simple, all in all.

I was afraid that someone would shout that my algorithm is much too
complicated. It is a great pleasure for me to know that the opposite
is acctually the case and that it can be easily implemented in much
less time than what I estimated. I don't have in mind that users
(the circle of persons that probably would use my code) should do
any optimization. My code does not run 'impressively' fast at all.
But, as I said, they don't normally need very high processing speed.
Finally, I don't agree that simplicity is a sin of an algorithm.

>
> > Therefore I have called it instead a
> > layman's data encryption algorithm, since it is designed by a layman
> > for use by laymen.
> >
> This scares me more than a little. Why should I - and I count myself a
> layman rather than a cryptographer, because while I have a fair deal of
> implementation experience, my mathematical background is much poorer than I
> would like - trust a cipher designed by a layman and not extensively
> analyzed over a cipher designed by a professional and analyzed by the
> cryptographic community? Why should anyone?

How can a brand new algorithm get 'extensively' analysed at the time
it comes out? The case is similar to that, say, when a new product
is put on the shelves of a supermarket. If nobody likes it, it
will disappear in relatively short time. It will stay, if it can find
sufficient customers.

>
> > (Hardcore professionals with their insistance on
> > rigorous mathematical proofs and the regulating officials taking
> > advices from them presumably would not have the least motivation to
> > examine, let alone to actually use, anything that is WEAK by name.)
> >
> I don't think that it's the WEAK aspect of the name, although that probably
> doesn't help. What deterred me from even looking at it was my inability to
> find an algorithm specification on your page - you seemed to only have the
> Fortran reference implementation. A reference implementation is not in any
> way a substitute for a well-written specification describing design goals,
> the algorithm in pseudocode at a high level (for those of us who have no
> desire to try to piece out Fortran) and describing any attacks that have
> been made or can be made on the cipher.
>
> This doesn't strike me as an insistence on rigourous mathematical proofs,
> but simply a basic requirement for any kind of understanding of the cipher.
> You've given absolutely no reason that I, as an implementor, would want to
> use your encryption scheme, and you've given no easy way for me, as an
> analyst, to even understand what your scheme is doing.
>
> If I just missed the specification, then please take my apologies, let me
> know where to find it, and I'll be glad to take a quick look over the scheme
> and see if I notice anything.

Since you claimed that my algorithm is very simple (which is in my
opinion true), I suppose that you have already read and understood it.
In the comments to the code I have explained very shortly what
is done in each round and have referred the issue of pseudo-random
generation to another article. With that, particularly since you
are implementor and therefore have barely difficulties in reading
codes (even in case you don't program in Fortran) and since I guess
you have much more programming experience than I (see above for
the difference in estimates of time needed to implement my algorithm)
I can't imagine that my well commented code presents you any real
difficulties at all. If there any concrete questions please kindly
let me know.

M. K. Shen

• Next message: William H. Geiger III: "Re: MD5 partial collisions"
• Previous message: Mok-Kong Shen: "Re: DES cracker built"
• In reply to: Bruce Schneier: "DES cracker built"