Issue #1689 has been updated by alexh.
We should stay away from non-reviewed work in this area. Bob Jenkins' IBAA/ISAAC implementations have been used and reviewed a fair number of times, unlike rcarey's work. The main reason we still have IBAA around is as a sort of safety net against any potential (implementation) bugs in the new CSPRNG. As such, it sounds like a pretty bad idea to try and tweak it like this. IMO this should be closed, but I'll leave it up to you. ---------------------------------------- Bug #1689: IBAA recommendation http://bugs.dragonflybsd.org/issues/1689#change-12535 * Author: robin.carey1 * Status: Feedback * Priority: Low * Assignee: * Category: Crypto * Target version: 4.2.x ---------------------------------------- Dear DragonFlyBSD bugs. DragonFly uses Bob Jenkins' IBAA CSPRNG for /dev/random. I have been doing some work on this CSPRNG, which is available from: http://www.leopard.uk.com/IBAA64 I have two recommendations to make with regard to the IBAA CSPRNG which is used in DragonFly: (1) Use my rounded up BETA=32/SHIFT=20 values as opposed to the original SHIFT=19 value in the original IBAA CSPRNG/algorithm (details available from above URL) and (2) Use a 32-bit counter. This is a minor modification, which gurantees no bad states (an initial state of all zeroes maps back to an internal state of all zeroes after 256 iterations), and also guarantees a cycle length of at least 2^32. Details on how to implement the counter are available from the above URL. It's a minor modification which has minimal performance impact. Also, a 64-bit version of IBAA is available from the above URL. Matthew Dillon seemed to be interested in that, with regard to the 64-bit x86-64 port of DragonFly (but I haven't heard back from him about that). ---Files-------------------------------- unnamed (1.4 KB) -- You have received this notification because you have either subscribed to it, or are involved in it. To change your notification preferences, please click here: http://bugs.dragonflybsd.org/my/account