Re: Make things interesting

From: Russ Lyttle (lyttlec_at_earthlink.net)
Date: 06/28/03 

Date: Sat, 28 Jun 2003 00:10:40 GMT

Tom St Denis wrote:

> Russ Lyttle wrote:
>> Tom St Denis wrote:
>> <<SNIP for space>>
>> Perhaps Tom, contact, and others could give the following code a try. Its
>> a bit long, but I think it would be interesting to see how it runs on
>> your tests. It should be a fairly faithful reproduction of Algorithm M
>> and might turn into something useful.
>
> Yours uses a LCG as the underlying PRNG. This has been proven to be
> weak. No joy.
>
> Tom
Correct. But the post is only intended to test the code technique. i.e. are
the results at least as good as the PRNG? I see lots of code that tries
cute tricks and results in less secure output than simple by-the-book
implementations. I'm trying a lot of cute tricks in the sample. It would be
nice to compare the stats of the output to the stats of the base PRNG.
Notice that the PRNG is easily replacable with anything you wish. In real
code, I would use two different PRNGs, one for X and another for Y. It
would be advisable to have relative prime periods for the PRNGs if
possible. In all, there is a big jump from the code posted to working code.
Question is, is the effort worth while?
BTW, the runtime cost for filling the S array is 0. Everything is calculated
at compile time. g++ really bogs down if the S array gets much bigger than
1024. I'm not a big fan of C++ templates, but having the compiler do that
much computation is interesting! 

-- 
Russ Lyttle
Not Powered by ActiveX

Relevant Pages

  • Re: sci.crypt sandbox?
    ... "Tom St Denis" wrote ... before being used to seed the pring used to encrypt the file. ... recover the entire prng just before the file is encrypted), ... My hash routine was designed to generate a *unique* hash for small ...
    (sci.crypt)
  • Re: BitBox PRNG
    ... | And comments to Tom St. Denis: ... So why didn't you analyze your design, write a formal proposal and then ... | 2) I'm not suggesting that this PRNG be used in any particular ...
    (sci.crypt)
  • Re: new /dev/random
    ... Tom St Denis writes: ... ]> Another tactic is to keep making unsupported statements until the other ... what everyone else has been calling a PRNG. ... It makes an attempt to be a RNG. ...
    (sci.crypt)
  • Re: strengthening /dev/urandom
    ... > Tom St Denis wrote: ... >> these theoretical problems and one PRNG that has them. ... >> if the cost isn't that high? ...
    (sci.crypt)
  • Re: Reinitialising random number generator
    ... the PUT array in different ways, even if the size happens to be the same. ... treated by the call is up to the implementor. ... routine init_random_seed) is standard Fortran 95. ... - The quality and variation of the implementation of the PRNG. ...
    (comp.lang.fortran)