# Dynamics based Privacy Preservation in Decentralized Optimization

arXiv: https://arxiv.org/abs/2207.05350 · HTML: https://mlsanity.com/paper/2207.05350/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Jul 2022.

## Built on by (most cited first, 5 of 5)

- [On the Tradeoff between Privacy Preservation and Byzantine-Robustness in Decentralized Learning](https://mlsanity.com/paper/2308.14606/) (2023)
- [Locally Differentially Private Gradient Tracking for Distributed Online Learning over Directed Graphs](https://mlsanity.com/paper/2310.16105/) (2023)
- [Privacy-Preserving Distributed Optimization and Learning](https://mlsanity.com/paper/2403.00157/) (2024)
- [Locally Differentially Private Distributed Online Learning with Guaranteed Optimality](https://mlsanity.com/paper/2306.14094/) (2023)
- [A Weighted Gradient Tracking Privacy-Preserving Method for Distributed Optimization](https://mlsanity.com/paper/2509.18134/) (2025)

## Abstract (arXiv)

With decentralized optimization having increased applications in various domains ranging from machine learning, control, sensor networks, to robotics, its privacy is also receiving increased attention. Existing privacy-preserving approaches for decentralized optimization achieve privacy preservation by patching decentralized optimization with information-technology privacy mechanisms such as differential privacy or homomorphic encryption, which either sacrifices optimization accuracy or incurs heavy computation/communication overhead. We propose an inherently privacy-preserving decentralized optimization algorithm by exploiting the robustness of decentralized optimization to uncertainties in optimization dynamics. More specifically, we present a general decentralized optimization framework, based on which we show that privacy can be enabled in decentralized optimization by adding randomness in optimization parameters. We further show that the added randomness has no influence on the accuracy of optimization, and prove that our inherently privacy-preserving algorithm has R-linear convergence when the global objective function is smooth and strongly convex. We also rigorously prove that the proposed algorithm can avoid the gradient of a node from being inferable by other nodes. Numerical simulation results confirm the theoretical predictions.