Optimising Privacy-Preserving Computations

There are different techniques which enable privacy-preserving computations, for example homomorphic encryption, multi-party computation and functional encryption. Two of these techniques are discussed in this thesis: homomorphic encryption (HE) and multi-party computation (MPC). Homomorphic encryption enables a third party to perform computations on encrypted data and therefore enables one to outsource computations on sensitive data to an untrusted party without compromising the privacy of the data. Multi-party computation allows several mutually distrusting parties to jointly compute a function over their inputs without revealing those inputs to other parties. This thesis aims at increasing the performance of these privacy-preserving computation techniques by improving generic operations and designing tailored solutions for specific application scenarios. Our research shows privacy-preserving computation techniques can solve privacy issues in real-life application scenarios by designing efficient solutions for a wide range of application settings: electricity forecasting, genome-wide association studies, disease prediction, fraud detection, substring search and threshold signatures. However, designing these solutions with the current state-of-the-art schemes requires a cumbersome selection of optimisation techniques and parameters.