New directions in secure computation

Conventional Machine Learning and Cloud Computing solutions require access to the raw data which generates potential security and privacy risk. Homomorphic Encryption method and Secure Multi-Party Computation protocol are both suitable solutions for security vulnerability. Homomorphic Encryption and Secure Multi-Party Computation provide to perform arbitrary computation on encrypted data without revealing the data and give the result of the same computation performed on unencrypted data. In this thesis, the essential concepts, mathematical background, property of Homomorphic Encryption schemes, practical implementations, state of the art Cloud approaches, current industry and potential applications for Machine Learning algorithms are analyzed in detail.