Implementing ML Algorithms with HE

An increase in cloud-based computing leads to an increased worry in the security of user data. Typically, data is sent to a third-party server which performs analytics or machine learning on the data. However, in most of these scenarios, the data involved is sensitive and should remain private. Homomorphic encryption, a form of encryption that allows functions to be performed on encrypted ciphertext, allows privacy-preserving data analysis of existing large private, sensitive data sets. We implement and analyze the performance of linear regression and K-means clustering using the homomorphic encryption library SEAL and provide an extension of the SEAL library to matrix operations.