Re: Fixing RNG in Microsoft Windows?

From: Yama (Yama@yomama.com)
Date: 05/17/02 

From: Yama@yomama.com (Yama)
Date: Fri, 17 May 2002 05:28:32 GMT

On 16 May 2002 18:54:35 -0700, chudel@carolina.rr.com (chu) wrote:

>Yama@yomama.com (Yama) wrote in message news:<3ce2b1ef.61131131@news-central.giganews.com>...
>> On 15 May 2002 09:59:04 -0700, chudel@carolina.rr.com (chu) wrote:
>
>> >Thanks for this good answer. I wonder if this may be part of the
>> >OLEInitiatlize calls (do seed the RNG multiple times) which in part
>> >may be part of a createuuid/guid? In any case, it's only a guess if
>> >the crypto program is using this as the random seed or getting it from
>> >another source.
>> >
>> >Thanks again,
>> >./Chu
>>
>> You are welcome.
>>
>> I have a lot of interest in this part of the MS Crypto API (CAPI),
>> particularly the CryptGenRandom call.
>>
>> I have had no luck whatsoever determining the strength or entropy of
>> this call or the PRNG in general.
>>
>> If you or anyone has any info on the relative entropy of this PRNG
>> (without seeding manually) I would be very interested.
>>
>> Also, I'd be very interested to know if I make the CryptGenRandom call
>> and seed it with a very low entropy value number, does this weaken the
>> call to the entropy of the my seed or merely increase the strength by
>> an extremely small factor (the entropy value of the call + the entropy
>> value of my seed)?
>>
>This just shows my ignorance, but I'm happy to learn from the smarter
>sci.crypt community. I would have thought you could just as easily
>use something like CoCreateGuid that generates a "universally unique"
>id number and then take an md5 or sha hash of that - shouldn't that be
>quite random?
>
>For my entropy tests, I run a program called "ent"
>(http://www.fourmilab.ch/random/). Although to be honest, in my (very
>limited poor and weak) cryptanalysis experience, I'm really just
>looking to verify if something is "mostly random", not truely so. :)
>
>Cheers,
>./Chu
>
>c h u d e l - at - b e l l s o u t h -dot- n e t

I learn a lot here, too. That is the beauty of this place to me.

Plus I get to share, occasionally.

The GUID and then hash-it will be pretty random, and easy to create,
but I don't think it has a large enough degree of difficulty to
reproduce it, for it to be crypto-strength random, unfortunately.

But it is a neat idea. And if you are seeking merely random-looking
data, I would think your method is useful.

Relevant Pages

  • Re: A secure, trustworthy Win XP compitable encryption program with GUI interface?
    ... >> that most cryptographic hashes are even cryptographically strong. ... Also it is a fundamental property that any strong crypto ... the requirements of an entropy distillation function. ... CRC output stream and use it as an AES key to encrypt the CRC output ...
    (sci.crypt)
  • Re: new /dev/random
    ... There was none in the original version, which gave it the exact same ... reading more data than is in the entropy pool. ... if the crypto primitives ...
    (sci.crypt)
  • Re: strengthening /dev/urandom
    ... The Linux /dev/random mixer has some general-purpose entropy estimation ... /dev/random is safe, if you want a true RNG (i.e., true entropy, secure ... against information-theoretic attacks) or if you don't trust the crypto. ...
    (sci.crypt)
  • Re: Fixing RNG in Microsoft Windows?
    ... >> If you remove the seed value under the RNG key, ... >> crypto operation, ... If you or anyone has any info on the relative entropy of this PRNG ...
    (microsoft.public.win2000.security)
  • Re: Key entropy, stream entropy, block entropy, block population entropy AKA uniique stream length
    ... I understand some math and crypto but you've strayed far off those ... cannot get entropy from a source that does not exist. ... discredit something they want to bury. ...
    (sci.crypt)
Quantcast