Privacy optimization method for federated learning based on evolutionary game theory
Source abstract
Abstract
In distributed collaborative modeling, federated learning struggles with privacy-performance trade-offs, worsened by non-IID client data complicating convergence. Traditional fixed privacy budgets and static clipping, lacking adaptability to data heterogeneity, cause inefficiency and high leakage risks. We propose an evolutionary game-based optimization for federated learning: using Rényi entropy to quantify data diversity for differentiated privacy, modeling budget allocation as a multi-agent dynamic game for adaptive adjustment, and linking client contribution with budget via positive feedback to enhance participation and data quality. MNIST and CIFAR10 experiments show that it improves accuracy and robustness under differential privacy, optimizing the privacy-performance balance.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Privacy-Preserving Technologies in Data · Mobile Crowdsensing and Crowdsourcing · Privacy, Security, and Data Protection
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Information Science and Technology, Yunnan Normal University 1 , Kunming 650500,
College of Physics and Information Engineering, Zhaotong University 2 , Zhaotong 657000,
College of Computer Science, Sichuan University 3 , Chengdu 610065,
Funding
National Natural Science Foundation of China · Science and Technology Plan Project of Yunnan Province
References (26)
Show source-provided references
1. GFL-ALDPA: A gradient compression federated learning framework 352 based on adaptive local differential privacy budget allocation10.1007/s11042-023-16543-y ↗
2. A survey of differential privacy techniques for federated learning10.1109/access.2024.3523909 ↗
3. 2017 · 1273
4. DRL-based federated self-supervised learning for task offloading and resource allocation in ISAC-enabled vehicle edge computing10.1016/j.dcan.2024.12.009 ↗
5. Membership inference attacks against machine learning models
6. Exploiting unintended feature leakage in collaborative learning
7. J. Geiping , F.Gaede, H.Bauermeister, and M.Moeller, “Fast convex relaxations using graph discretizations,” arXiv:2004.11075 (2020).10.5244/C.34.157 ↗
8. Secure multi-party computation with secret sharing for real-time data aggregation in IIoT10.1016/j.comcom.2024.06.002 ↗
9. Privacy preserving machine learning with homomorphic encryption and federated learning10.3390/fi13040094 ↗
10. The algorithmic foundations of differential privacy10.1561/0400000042 ↗
11. A stackelberg incentive mechanism for wireless federated learning with differential privacy10.1109/lwc.2022.3181509 ↗
12. Advances and open problems in federated learning10.1561/2200000083 ↗
13. Deep learning with differential privacy
14. R. C. Geyer , T.Klein, and M.Nabi, “Differentially Private federated learning: A client level perspective,” arXiv:1712.07557 (2017).
15. Federated learning with differential privacy: Algorithms and performance analysis10.1109/tifs.2020.2988575 ↗
16. Evolutionary model owner selection for federated learning with heterogeneous privacy budgets
17. Adaptive privacy budget allocation in federated learning: A multi-agent reinforcement learning approach
18. H. Zhou , “Research on incentive mechanism and privacy protection scheme based on federated learning,” M.Sc. thesis, Guizhou University, Guiyang, China, 2023.
19. A survey of incentive mechanism design for federated learning10.1109/TETC.2021.3063517 ↗
20. A game-theoretic framework for incentive mechanism design in federated learning
21. DIM-DS: Dynamic incentive model for data sharing in federated learning based on smart contracts and evolutionary game theory10.1109/jiot.2022.3191671 ↗
22. A robust game-theoretical federated learning framework with joint differential privacy10.1109/tkde.2021.3140131 ↗
23. Computation and simulation of evolutionary game dynamics in finite populations10.1038/s41598-019-43102-z ↗
24. Cloud Federation formation using genetic and evolutionary game theoretical models10.1016/j.future.2019.10.008 ↗
25. Evolutionary model owner selection for federated learning with heterogeneous privacy budgets
26. Differential privacy in federated learning: An evolutionary game analysis10.3390/app15062914 ↗