A Distributed Privacy-Preserving Algorithm using Row Stochastic Weight Matrix with Locally Balanced Noise
Ziye Liu, Wei Wang, Fanghong Guo · 2023
Privacy preservation for distributed optimization algorithms in multi-agent systems has warranted widespread concern in recent years due to the urgent security requirements. In this paper, a novel distributed subgradient privacy-preserving algorithm is proposed, where the weight matrix is row stochastic. The proposed algorithm holds an advanced noise-adding mechanism, where locally balanced noise is adopted to mask the real data that agents transmit to neighbors. It is shown that all the agents converge to an optimal solution of the optimization problem, while the subgradients and cost functions are preserved simultaneously.