computablelabs / hepy Goto Github PK

Based on feedback sourced from Twitter, it looks like using conda is best practices for dealing with mixed Python/C++ code bases. We should swap this repo to using conda for its build.

The Repo has started to get a little messy. Let's use this issue to track clean-up changes we should make.

• Clean up setup.py's user field as noted in comments on #11

There's a lot of boilerplate to generate when writing even simple homomorphic programs. Being able to experiment in Python would make the turnaround time on simple experiments much faster. Some simple Googling reveals that SWIG seems to be a good option for wrapping C++ libraries. Implement a functional SWIG wrapper for the classes here.

http://intermediate-and-advanced-software-carpentry.readthedocs.io/en/latest/c++-wrapping.html

Need to add an installation command as specified here: homenc#201

Having written a couple homomorphic programs, some paradigms are starting to emerge. Here are the basic phases:

• Set parameters of homomorphic scheme. These are currently set by user input flags at the top of the scripts.
• Construction of FHEcontext object to store scheme parameters.
• Construction of public key object FHEPubKey. Construction of private key object FHEPrivateKey.
• Encryption of data. This involves construction of EncryptedArray object which stores encrypted data.
• Actual homomorphic computation. This may involve some type fiddling to make sure that various ciphertexts are of the right type.
• Decryption followed by check to make the the computation was performed correctly.

I think that the set-up work can be factored out into a helper file that sets sensible defaults. The other phases can also likely be abstracted some to help cleanup.

As development continues on this repo, it would be good to get the tests running on Travis CI so we know that we're not accidentally breaking core functionality.