Party Protocol

Party protocols, encompassing secure multi-party computation (MPC), aim to enable collaborative computation on sensitive data without revealing individual inputs. Current research focuses on improving the efficiency of these protocols, particularly for machine learning tasks, using techniques like three-party models and novel privacy mechanisms such as the Poisson Binomial mechanism. This work addresses limitations in existing two-party protocols and explores efficient solutions for various applications, including federated learning and secure inference for transformer models, thereby enhancing privacy in data-driven applications. The development of more efficient and practical party protocols is crucial for advancing privacy-preserving machine learning and other sensitive data analyses.