Multi Party Computation

Secure multi-party computation (MPC) enables multiple parties to jointly compute a function on their private data without revealing individual inputs. Current research focuses on improving the efficiency of MPC for machine learning tasks, particularly with deep learning models like Transformers and diffusion models, addressing challenges such as high communication overhead and computationally expensive non-linear functions through techniques like quantization and optimized approximations of activation functions. This work is crucial for enabling privacy-preserving machine learning in sensitive domains like healthcare and finance, mitigating data poisoning attacks, and facilitating collaborative data analysis without compromising confidentiality. The development of robust and efficient MPC frameworks is driving progress in both theoretical computer science and practical applications of privacy-enhancing technologies.