Privacy Enhancement Technologies (PET) have gained attention in domains where privacy is essential, such as health analysis, financial analysis, machine learning, big data, etc. Homomorphic Encryption (HE) and secure multi-party computation (MPC) are two major approaches. The advances in research on these technologies were mostly at the theoretical level for many years. However, the focus shifted to their practical deployments and performance improvements during the last decade. In particular, research on the MPC side has been accelerated as it is faster than fully homomorphic encryption (FHE). MPC consists of several mutually distrusted parties, which hold a share or secret of each private value, and jointly evaluate a function without leaking any information other than the respective output.
While there have been notable advances in improving MPC performance, there are still several challenges ahead, not only related to further efficiency improvements but also in other aspects that are critical for their broad adoption, such as robustness, correctness verification, and management, including initial MPC setup and maintenance. For instance, there is a lack of solutions that handle real-world events such as participants joining or leaving the system, the coordination among clients and MPC service providers, handling failures, and more.
This project focuses on enhancing the MPC efficiency from the networking and management perspectives and proposing a protocol that automates the MPC system setup and management without degrading the efficiency. Some features include use of blockchain communication channel, participants registration (e.g., MPC server, input clients), authentication among participants, orchestration of MPC tasks, correctness verification, output delivery, and resumption of computation in case of node failures.
This project is supported in part by the US National Science Foundation and the US Air Force.
People: Oscar Bautista, Dr. Mohammad Hossein Manshaei, Richard Hernandez, and Dr. Kemal Akkaya
- O. Bautista, K. Akkaya and S. Homsi, “Outsourcing Secure MPC to Untrusted Cloud Environments with Correctness Verification,” 2021 IEEE 46th Conference on Local Computer Networks (LCN), 2021, pp. 178-184, doi: 10.1109/LCN52139.2021.9524971.
- O. G. Bautista and K. Akkaya, “Network-Efficient Pipelining-Based Secure Multiparty Computation for Machine Learning Applications,” 2022 IEEE 47th Conference on Local Computer Networks (LCN), 2022, pp. 205-213, doi: 10.1109/LCN53696.2022.9843372.