News for hackers. Stuff that matters.
Re:Specific Problems with Other RNGs by ahh on Wednesday November 11, 2015 @11:16AM (#10547368)
An interesting case would be something like hash flooding prevention. The cost of a full CSPRNG is probably not worth it--no security is broken if we lose the secrecy of one generated seed--but at the same time we want something that is at least nontrivial for an attacker to predict.

I agree that most cases you either don't care at all or need a real CSPRNG.

Re:Specific Problems with Other RNGs by tedunangst on Wednesday November 11, 2015 @11:22AM (#10547399)
Preventing hash flooding requires a single 128 bit seed per table. How expensive is it to generate that with a real CSPRNG?
Re:Specific Problems with Other RNGs by ahh on Wednesday November 11, 2015 @11:51AM (#10547577)
Depends how many hash tables you have! :)

I agree that it's a marginal gain here, but Thomas asked for a case where predictability of "more difficult than a LCG or Mersenne Twister, but not infeasible under crypto assumptions" would have value. This would be an example.