Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.29845 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914614199975936 |
|---|---|
| author | Zhang, Zhiguo Kui, Cheng Ma, Qian Wu, Dongrui |
| author_facet | Zhang, Zhiguo Kui, Cheng Ma, Qian Wu, Dongrui |
| contents | Distributed online stochastic optimization has received extensive attention in large-scale distributed learning and other related fields due to its unique advantage in processing streaming data. However, information exchange through the communication network during the optimization process may lead to privacy leakage. To address this issue, this paper proposes a locally differentially private distributed online stochastic optimization algorithm that employs an elaborately designed dynamic stochastic quantizer to mask the exchanged information prior to communication. Theoretical analysis shows that the proposed algorithm not only converges almost surely to the optimal solution but also achieves $(0,δ^i)$-local differential privacy for each agent $i$ even when the number of iterations tends to infinity. Furthermore, the algorithm is fully distributed and applicable to scenarios where the interaction network among agents is a directed graph. To the best of our knowledge, this is the first work on distributed online stochastic optimization that simultaneously achieves exact convergence and rigorous local differential privacy over a directed graph by exploiting quantization effects. Numerical experiments of distributed online training on the mushroom classification dataset, handwritten digits recognition dataset, and brain-computer interface dataset verify the effectiveness of the proposed method. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_29845 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Local Differential Privacy via Dynamic Quantization in Distributed Online Stochastic Optimization Zhang, Zhiguo Kui, Cheng Ma, Qian Wu, Dongrui Optimization and Control Distributed online stochastic optimization has received extensive attention in large-scale distributed learning and other related fields due to its unique advantage in processing streaming data. However, information exchange through the communication network during the optimization process may lead to privacy leakage. To address this issue, this paper proposes a locally differentially private distributed online stochastic optimization algorithm that employs an elaborately designed dynamic stochastic quantizer to mask the exchanged information prior to communication. Theoretical analysis shows that the proposed algorithm not only converges almost surely to the optimal solution but also achieves $(0,δ^i)$-local differential privacy for each agent $i$ even when the number of iterations tends to infinity. Furthermore, the algorithm is fully distributed and applicable to scenarios where the interaction network among agents is a directed graph. To the best of our knowledge, this is the first work on distributed online stochastic optimization that simultaneously achieves exact convergence and rigorous local differential privacy over a directed graph by exploiting quantization effects. Numerical experiments of distributed online training on the mushroom classification dataset, handwritten digits recognition dataset, and brain-computer interface dataset verify the effectiveness of the proposed method. |
| title | Local Differential Privacy via Dynamic Quantization in Distributed Online Stochastic Optimization |
| topic | Optimization and Control |
| url | https://arxiv.org/abs/2605.29845 |