Privacy-Enhanced Distributed Event-Triggered Algorithm for Constrained Optimization with Shared Constraints

doi:10.1109/tac.2026.3731925

Privacy-Enhanced Distributed Event-Triggered Algorithm for Constrained Optimization with Shared Constraints

2026-09-07

0

PRE

AI

DOI：10.1109/tac.2026.3731925

Original

Original request for help

Save

Abstract

En 中文

This paper studies privacy-concerned distributed optimization problems with shared inequality and equality constraints, in which each agent has access only to its local cost and constraint functions. To address privacy concerns, we inject noise into the exchanged state variables against potential attackers. We propose an event-triggered differentially private distributed algorithm that ensures both accurate convergence and rigorous $\epsilon$-differential privacy over infinite iterations. We demonstrate the effectiveness of the event-triggered mechanism for privacy protection, and mathematically prove that it reduces the privacy budget, thereby enhancing the privacy level. Moreover, the developed algorithm does not require the strong convexity of local cost functions or boundedness of their gradients. We also propose a unified analytical approach that can prove the convergence and privacy properties under both constant and diminishing step-sizes, and provide a theoretical analysis of the trade-off between convergence accuracy and privacy level. Finally, numerical simulations are conducted to evaluate the performance of the proposed algorithm.

Keywords:

Distributed optimization

Differential privacy

Shared inequality constraints

Event-triggered mechanism

Privacy-Enhanced Distributed Event-Triggered Algorithm for Constrained Optimization with Shared Constraints

Abstract

Journal

Organization

Cited Papers