tssci security

Testing for randomness and predictability using Burp Sequencer

Sorry I haven't posted in forever. Dre's been covering for me while I've been super busy with finishing up school, reading, work, and other projects. I think Dre's packed more information in the last month than I did all year. 2007 Security Testing Tools in Review alone is worth a third or fourth reading.

Anyways, here's one of the things I've been looking at lately. Portswigger released a new version of the Burp Suite with several new tools included, one of which is called "Sequencer." What Burp Sequencer is, is a session token analyzer that tests the degree of randomness and predictability in a sample data set. From the Burp Sequencer help:

Burp Sequencer employs standard statistical tests for randomness. These are based on the principle of testing a hypothesis against a sample of evidence, and calculating the probability of the observed data occurring, assuming that the hypothesis is true:

Burp Sequencer can run against a live capture or copy/paste loaded strings. When browsing a site with Burp acting as a proxy, you can send requests to Sequencer using the "action" button for a live capture of session tokens. I began playing with this tool to see how "random" the session tokens are in various sites, and then decided to run it against a couple PRNG's and compare them to others in an attempt to find the best source of entropy data to use for creating session tokens.

NIST has published over the years several requirements and guidelines for creating and testing random number generators as part of their cryptographic toolkit and under the FIPS 140-2, Security Requirements for Cryptographic Modules. Burp Sequencer conducts a series of full FIPS tests for randomness, among others including spectral, correlation and compression tests. More information on FIPS 140-2 standard is available at the Cryptographic Module Validation Program (CMVP) site.

Random.org has been around since 1998, offering an online, "true" random number generator whose source of entropy is atmospheric noise picked up by radio receivers. You can read more information about it from their FAQ. What's cool, is Random.org publishes performance statistics available on a daily basis.

I compared Linux's /dev/urandom to Random.org. My sample set met the following requirements:

  1. 16 character length
  2. 10,000 tokens
  3. Alphanumeric
  4. Repeats allowed

To generate 10,000 tokens using /dev/urandom, I issued the following command:

$ cat /dev/urandom | tr -cd [:alnum:] | fold -w 16 | head -10000

I ran the sample sets from /dev/urandom and Random.org through Burp Sequencer and concluded the following:

The overall quality of randomness in /dev/urandom is considered to be poor, with a significance level of 1% and the amount of effective entropy to be 29 bits. Random.org fared worse, with an amount of effective entropy to be only 25 bits.

I've made my sample data sets available online for anyone to download, from /dev/urandom and random.org results.

Some other random number generators that I encourge you analyze on your own using Burp Sequencer include QRBGS (a quantum-RNG), LavaRnd (a RNG that uses lava lamps for entropy), and HotBits (uses radioactive decay as a source of randomness).

Posted by Marcin on Friday, December 21, 2007 in Security.

blog comments powered by Disqus
blog comments powered by Disqus