I am working on analyzing the "made-up-ness" of company's financial reports by running statistical tests on the numbers in their SEC filings, using the following assumptions:
Humans are bad at making up random numbers.
Crooks are lazy.
Therefore, crooks making up numbers are unlikely to use a high quality random number generator to generate their numbers and so companies practicing financial fraud will have noticeable deviations from expected randomness.
I hope the end result will be a single number that can be used to rank companies relative to each other in the "made-up-ness" of their results.
Old stuff
I finished the project below; for results see:
http://www.nmt.edu/~val/review/ext2fsck.pdf
I am working on improvements to ext2 - specifically, trying to reduce fsck time while preserving high performance, robustness, and especially simplicity of implementation.
My home page:
http://www.nmt.edu/~val
If you know file systems, come talk to me!