RNG: is it possible to spoil /dev/random by seeding it from (evil) TRNGs

[Christoph Anton Mitterer <calestyo@scientia.net>]

Hi.

This is a question towards the crypto/entropy experts.

When seeding the kernels entropy cache (which is then ultimately used
for /dev/random), e.g. by (semi-)TRNGs like haveged[0],
audio-entropyd[1], Simtec’s Entropy Key[2] or friends... can one spoil
the randomness by that or is this impossible by design?

Of course it's easy to check the distribution of these randomness
sources, but as we see on the plain Mersenne Twister, a "perfect"
distribution is not necessarily usable for cryptography.


Further, one could imagine that closed products like the Entropy Key are
hacked or have backdoors, which may make them produce subtle patterns
that could later be used in cryptoanalysis.
(This is in no way a claim, that Simtec would do this,... just an
example.)


Cheers,
Chris.



[0] http://www.issihosts.com/haveged/
[1] http://www.vanheusden.com/aed/
[2] http://www.entropykey.co.uk/