I would appreciate any ideas for applying machine learning methods and algorithms in cryptography problems.
I'm thinking about trying to find a collision in Hash algorithms
Neural Cryptography is an existing machine learning based Cryptographic approach. I feel it still have to be improved towards its robustness.
You may need to read about physical cryptography .
Please check this paper for Ronald L. Rivest, one of the best : http://people.csail.mit.edu/rivest/pubs/Riv91.pdf
I've always suspected there's an application of the blind source separation problem to discovering the plaintext for a known ciphertext and a characterized keystream. This would be more general Bayesian inference over a number of samples having the same plaintext (a crib) but different keystreams potentially with a complex function applied. I've used a simple version in my own cryptanalysis.
- Badges
- Science topic