Differential Privacy for Distributed Consensus with Partial Observations

Abstract

In a network, distributed privacy-preserving algorithms protect sensitive information by adding noise to the message before it is transmitted to other nodes in the network. Most existing works on differentially private distributed algorithm assumed that the adversary has access to all messages transmitted through the network. However, this is not a realistic assumption when dealing with real-world applications with large networks such as power networks. Very often the adversary can only breach a small part of the network and has limited observation of the messages. In this paper, we study how partial observations affect the privacy guarantee of a distributed consensus algorithm. For the dist..