Privacy-Preserving Average Consensus via Homomorphic Encryption

With increased security awareness and increased attack incidents, it is becoming increasingly important to protect the privacy of data during storage and communication. As the basis of distributed systems, consensus is widely used in automatic control, signal processing and optimization. In this paper, a consensus protocol based on partially homomorphic encryption is applied for undirected networks, which can achieve privacy-preserving average consensus. In this method, each agent can participate in information interaction with its neighbors using a virtual initial state which is generated based on the Paillier cryptosystem, thus protects the privacy of the initial state value. Different from the original cryptography-based methods, our method is fully decentralized without the help of any aggregators or third parties. Numerical simulation shows the effectiveness of this method and is compared with other methods.