Leo de Castro
MIT EECS | Lincoln Laboratory Undergraduate Research and Innovation Scholar
Efficient Two-Party Computation from Plaintext-Ciphertext Homomorphic Operations
- Theory of Computer Science
As cloud computing becomes more powerful, solutions for securely outsourcing computations on sensitive data become more necessary. Homomorphic encryption schemes have tremendous potential to provide analytical access to encrypted data while still maintaining privacy. Modern homomorphic schemes are fully functional, but have yet to be implemented in practical applications. Moreover, the asymptotic efficiency of homomorphic schemes is not well understood in the context of complex circuits, and comparisons can be difficult as different schemes have different potential optimizations through single-instruction multiple-data (SIMD) operations. This project will compare the efficiency of the BGV and FV homomorphic schemes when running three fundamental circuits for data sharing: ARGMAX, SETDIFF, and INTERSECTION, with the goal of creating an efficient suite of functions for secure genome analysis.
My SuperUROP project is in the field of lattice-based cryptography. Previous classwork and meetings with my professor (who is now my SuperUROP supervisor) led to my interest in this project. I hope to learn more about lattices and their cryptographic applications, as well as how to better communicate and share my results. I am excited to use these skills both in this project and in future work.