On: A Family of Better Random Number Generators

[plopilop]

The website does not seem to have been updated but the prediction security of PCG has been fully broken recently (and does not rely on the "Challenging" part of the spectrum): https://hal.inria.fr/hal-02700791/

Hence any cryptographic use of PCG is definitely prohibited

[Straw → plopilop]

Interesting that the strongest variant took an average 20,000 CPU-hours to break- so definitely not cryptographic strength, but not a walk in the park either.

[logos01 → Straw]

I'd call that a walk in the park if security is concerned.

[tptacek → Straw]

20,000 core-hours is, what, a couple hundred bucks?

[tptacek → plopilop]

This despite the extremely misleading green "Challenging" grade the authors gave PCG's security on their website.