Re: Good enough for crypto?

From: Scott Wilber (swilber_at_comscire.com)
Date: 12/04/03 

Date: 4 Dec 2003 14:02:28 -0800

r3769@aol.com (R3769) wrote in message news:<20031203182450.26118.00000099@mb-m06.aol.com>...
> Scott Wilber wrote:
>
> >The Mersenne Twister followed by our infinitely recursive,
> >multiple-feedback shift register (IRMFSR) stirring function produces
> >sequences that have passed all our tests cumulative to 10's of
> >trillions of bits. (By the way, NO other generator has passed our
> >tests so far)
> >
>
> I don't know what "pass" means, but I just fed 2.20E+9 bits of the following
> into your tests. Got a 29.0+ (whatever that means)..
>
> 10 p=2^35-31
> 20 i=0
> 30 n=1
> 40 s=0
> 50 n=p-1-((p*i-1)@n)
> 60 i=i+1
> 70 s=(s+n\256)@65536
> 80 random16bits=s
> 90 goto 50
>
> a@b==a mod b (6@2=0, 10@3=1)
> a\b==a div b (6\2=3, 10\3=3)
>
> Good enough for crypto? Hardly. Good enough to "pass" your tests? Probably.
> Let you know when I get to 40+...

I was not implying that any pseudorandom generator, like the one
described in the snip above, was "good enough for crypto." This
thread is primarily concerning the ComScire PCQNG TRNG: its
construction, entropy source and statistical properties.

The RNGmeter is designed to test a sequence until it fails at a
statistically significant level. If you get to 42+ that would make it
one of the best ones we have seen. 29.0+ is not impressive. 40+
would mean that by our testing algorithms, the sequence under test is
able to produce a sequence of at least 2**40 bits (about 1 trillion,
or more) without any statistically significant deviation. When the
little + disappears, that means the sequence has failed and the
statistically "good" length is 2**QF, where QF or quality factor is
the number indicated in the slider.

Please do let us know when that happens.

Scott

>
> Problems: (1) show p is prime iff n=1 for some i>0 and (2) find the period of
> n[i]. Have fun!
>
> rich

Relevant Pages

  • Re: Good enough for crypto?
    ... entropy source and statistical properties. ... >would mean that by our testing algorithms, the sequence under test is ... >or more) without any statistically significant deviation. ...
    (sci.crypt)
  • Re: Good enough for crypto?
    ... and Scott Wilber wrote: ... to which Mok-Kong Shen tried to clarify ... >> sequence used by the opponent is really generated, ... 'What is the per-bit entropy of a ...
    (sci.crypt)
  • Re: Good enough for crypto?
    ... > and Scott Wilber wrote: ... >> sequence being generated by a particular physical source?'. ... per-bit entropy of the output sequence. ... then the output sequence has exactly zero ...
    (sci.crypt)
  • Re: Good enough for crypto?
    ... > Scott Wilber wrote: ... The tester provides a quality factor that estimates the ... number of continuous bits in a sequence generated by any RNG that will ...
    (sci.crypt)
  • Re: Good enough for crypto?
    ... > Scott Wilber wrote: ... >> I suggest you go to the webpage ... >> sequence being generated by a particular physical source?'. ... See my response to "Bob" for your answer. ...
    (sci.crypt)
Quantcast