Bsides LV 2014 - Untwisting The Mersenne Twister: How I killed the PRNG - 05Aug2014

05 Aug 2014 - Bsides Las Vegas 2014
Joe "moloch" - Bishop Fox
Dan "AltF4" Petro - Bishop Fox
www.bishopfox.com
www.bishopfox.com/blog/2014/08...
www.irongeek.com/i.php?page=vi...
Untwisting The Mersenne Twister: How I killed the PRNG
Applications rely on generating random numbers to provide security, and fail catastrophically when these numbers turn out to be not so “random.” For penetration testers, however, the ability to exploit these systems has always been just out of reach. To solve this problem, we’ve created “untwister:” an attack tool for breaking insecure random number generators and recovering the initial seed.
We did all the hard math, so you don't have to! Random numbers are often used in security contexts for generating unique IDs, new passwords for resets, or cryptographic nonces. However, the built-in random number generators for most languages and frameworks are insecure, leaving applications open to a series of previously theoretical attacks. Lots of papers have been written on PRNG security, but there's still almost nothing practical you can use as a pentester to actually break live systems in the wild. This talk focuses on weaponizing what used to be theoretical into our tool: untwister. Let's finally put rand() to rest.
DISCLAIMER: This video is intended for pentesting training purposes only.

[@Houshalter]

There are a few, very important, minutes of video missing at 21 minutes.

[@anteconfig5391]

Can I ask about why the video seems cut at the beginning and at the 21 min mark?

[@tiernanmorgan]

if you know seed is datetime could you use this to reverse enginer a slot machine?

[@weedaq]

what is this supposed to be?

[@Omnifarious0]

Mersenne Twister is a recipe computers follow to generate a series of numbers that seem to be random. In truth, the numbers are not at all random. And, in fact, after you see a few numbers in the series, it's not that hard to guess what the rest of the numbers are going to be. That doesn't mean Mersenne Twister is bad. It's designed for games and monte carlo simulations (simulations of things that use statistical randomness for certain parts of the simulation). It was never designed to be used in cases where the predictability of the random numbers was a bad thing. From a statistical standpoint, it's very hard to tell the series of numbers generated by Mersenne Twister from a sequence of random numbers. It works very, very well for games and monte carlo simulations. The problem is, uninformed and sloppy programmers use it for generating things like session ids when you log into websites. If session ids are predictable, someone could hijack your session (i.e. start interacting with the website as if they had your password even when they don't) by guessing your session id. This video presents a tool that can be used to guess the rest of the numbers in the sequence given a few numbers from the sequence. This tool can be used to accomplish attacks of the kind I just described. The math involved in figuring this out is non-trivial. And while I haven't watched the video, other comments indicate that the people presenting the tool to not describe at all how it works which is very unsatisfying. The lack of explanation moves the tool closer to the realm of being purely a hacking tool instead of a tool for security researchers. After all, someone who just wants to break into things doesn't really care how the tool works, just that it does.

[@naninca]

@@Omnifarious0 good explanation

[@Omnifarious0]

@@naninca - I'm glad _someone_ appreciated it. :-)

[@keeper2828]

Incredibly poor presentation with near to no explanation whatsoever. "Here's a script, it cracks MT, and that's it". The presenters have no idea what's going on.