Privacy Preserving Recommendation System:
The scheme is secure against a malicious server, is independent of a trusted third party, and is efficient. It uses oblivious pseudo-random functions, secret sharing, homomorphic encryption, and traditional encryption protocols. It is implemented in C to accommodate the ed25519-donna package. Preliminary results show that our system is scalable on the MovieLens 1M dataset.
Atoms of Style:
Performed author attribution of binaries to see if elements of the original file can be used to identify its authors. We generated code property graphs of decompiled binaries using the Joern decompiler. We then trained the graphs on the Weka machine learning tool to predict the authors of test set binaries. Such techniques are used in applications such as malware detection, intellectual property theft.
Capability-based access control :
Implemented a system that does capability-based resource management, where cryptographically-verfied channels of communication were used between the entities passing the resource.
