Privacy Preserving Distributed Optimization via Paillier Encryption and Randomness Injection

With the rapid development of technologies such as embedded computing, wireless sensing and communication, distributed optimization has received increasing attention in the field of cyber-physical systems. Current research on distributed optimization is mainly about convergence performance analysis. With the wide application of distributed optimization in fields such as big data and cloud computing, the privacy protection of data plays a more and more crucial role in practical applications. To provide privacy protection against both honest-but-curious attackers and eavesdroppers, we propose a novel distributed optimization algorithm which embeds Paillier encryption and randomness into local interaction protocol of nodes. Different from differential privacy based approaches which sacrifice optimization for privacy protection, our approach is able to guarantee both the optimization accuracy and privacy preservation. The convergence performance and privacy protection performance are systematically analyzed, and simulations results are provided to verify the theoretical predictions.