Abstract:In recent years, there has been a growing interest in distributed optimization, which collaboratively attains an optimum by exchanging information with neighbours. Among the various distributed algorithms available, optimization with gradient tracking is particularly notable for its superior convergence results, especially in the context of directed graphs. However, privacy concerns arise when gradient information is transmitted directly which would induce more information leakage. Surprisingly, the literature has not adequately addressed the associated privacy issues. In response to the gap, our paper proposes a privacy-preserving distributed optimization algorithm with gradient tracking by adding noises to transmitted messages, namely, the decision variables and the estimate of the aggregated gradient. We prove two dilemmas for this kind of algorithm. In the first dilemma, we reveal that this distributed optimization algorithm with gradient tracking cannot achieve $ \epsilon$-differential privacy (DP) and exact convergence simultaneously. Building on this, we subsequently highlight that the algorithm fails to achieve $ \epsilon$-DP when employing non-summable stepsizes in the presence of Laplace noises. It is crucial to emphasize that these findings hold true regardless of the size of the privacy metric $ \epsilon$. After that, we rigorously analyse the convergence performance and privacy level given summable stepsize sequences under Laplace distribution since it is only with summable stepsizes that is meaningful for us to study. We derive sufficient conditions that allow for the simultaneous stochastically bounded accuracy and $ \epsilon$-DP. Recognizing that several options can meet these conditions, we further derive an upper bound of the mean error's variance and specify the mathematical expression of $ \epsilon$ under such conditions. Numerical simulations are provided to demonstrate the effectiveness of our proposed algorithm.

Differentially Private Gossip Gradient Descent

Gossip Gradient Descent.

A(DP)$^2$SGD: Asynchronous Decentralized Parallel Stochastic Gradient Descent with Differential Privacy

A(DP)$^2$2SGD: Asynchronous Decentralized Parallel Stochastic Gradient Descent with Differential Privacy

Locally Differentially Private Gradient Tracking for Distributed Online Learning over Directed Graphs

Differentially Private Nash Equilibrium Seeking in Quadratic Network Games

Tailoring Gradient Methods for Differentially-Private Distributed Optimization

An Accelerated Gossip-Based Distributed Gradient Method

Distributed Gradient Tracking for Differentially Private Multi-Agent Optimization with a Dynamic Event-Triggered Mechanism

Differentially Private Linear Regression over Fully Decentralized Datasets

Differentially Private Decentralized Deep Learning with Consensus Algorithms

Differential Privacy in Distributed Optimization with Gradient Tracking

Differentially Private Distributed Estimation and Learning

Stochastic gradient descent with differentially private updates

Differentially Private Distributed Resource Allocation Via Deviation Tracking

Differentially private stochastic gradient descent with low-noise

Distributed Dynamic Online Learning with Differential Privacy Via Path-Length Measurement.

Distributed Least-Squares Optimization Solvers with Differential Privacy

Differentially Private Learning with Small Public Data.

DPDR: Gradient Decomposition and Reconstruction for Differentially Private Deep Learning

Generalizing Differentially Private Decentralized Deep Learning with Multi-Agent Consensus