[tahoe-dev] about pycryptopp unit test

[Jack Lloyd]

On Thu, May 20, 2010 at 11:11:34PM +0800, yu xue wrote:
> Hello, everyone:
>     I am a student who is lucky to be choosed as one of this year's gsoc
> students. My name is Yu Xue. Next week the gsoc will start. My gsoc project
> is "100 year cryptography". This project's main purpose is to implement some
> combiners of cryptographic algorithms. The first part is combiner of
> ciphers. In this phase, I will write the documents and API, write test
> harness including unit tests, tests vectors etc, and implement the combiner
> of block cipher based on the test cases which is mainly comber of AES using
> CTR mode and XSalsa20. The second part is combiner of hash function---Comb4P
> which has three round feistel and the round function is composed of xoring
> of H0 and H1 with the help of round index i. The above two parts is the main

Yay!

Some things that need to be resolved:

 - Which hash functions will we combine? SHA-256 seems like an obvious
   candidate for one half, since it is relatively standard and thought
   to be secure at the moment, and is relatively efficient on 32-bit
   machines. But what of the other half? One of the SHA-3 candidates
   seems plausible, but which one? Ideally, we would choose precisely
   the function that will be eventually become SHA-3, but with 14
   round 2 candidates the odds of this are not terribly high.

   We could also instead combine SHA-256 with another hash already in
   Crypto++, for instance RIPEMD-160, Whirlpool, or SHA-512.

 - Will Comb4P be implemented in Python? In C++? Should we try to get
   Wei Dai to adopt Comb4P into the next version of Crypto++?

-Jack