Re: [GNU Crypto] Fortuna test results

[Casey Marshall <csm@xxxxxxx>]

On Oct 21, 2005, at 10:48 AM, jrandom-gnucrypto@xxxxxxx wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi all,
>
> I know the Fortuna implementation isn't production ready yet, but I
> just wanted to give a heads up as to some end user tests I ran the
> other day.  Basically, the data pulled from Fortuna isn't too
> random - a simple gzip test against the output allows substantial
> compression, while the java.security.SecureRandom in Sun's JVM or
> even java.util.Random does not:
>
> - -rw-r--r--  1 jrandom users  4652685 Oct 21 18:25 /tmp/testfortuna
> - -rw-r--r--  1 jrandom users 67129357 Oct 21 18:26 /tmp/testrandom
> - -rw-r--r--  1 jrandom users 67129357 Oct 21 18:26 /tmp/ 
> testsecurerandom
>
> The throughput of the Fortuna implementation is pretty good though,
> as I could pull 8-9MBps on my machine (p4 3ghz), spiked up to 21MBps
> when gcj'ed with -O3.
>
> Test code attached (javac -cp gnu-crypto-2.1.0.jar prng.java).
>
> I haven't had time to dig into why the output is off, but I'm not
> really an expert in prng implementations.  If there's something I'm
> doing wrong or if there's an easy fix, please, let me know, as I'd
> love to use Fortuna.  In any case, perhaps the attached could get
> turned into a unit test for PRNGs - if the file size written is less
> than 64MB, the PRNG isn't random.

The counter increment is wrong, and is just outputting the same value  
(a lot of 0's) over and over again, feeding that into the generator,  
which then produces a lot of encrypted zeros.

The incrementCounter function of the inner class Generator should be:

    private void incrementCounter()
    {
      for (int i = 0; i < counter.length; i++)
        {
          counter[i]++;
          if (counter[i] != 0)
            break;
        }
    }

This appears to fix the problem; thanks for the bug report!


_______________________________________________
gnu-crypto-discuss mailing list
gnu-crypto-discuss@xxxxxxx
http://lists.gnu.org/mailman/listinfo/gnu-crypto-discuss