A Novel Chaotic-Map-Driven Privacy Protection Mechanism for Federated Learning
Abstract
To address the persistent privacy leakage risks caused by sharing model parameters in federated learning, this paper proposes a dual-protection mechanism that integrates chaotic-map-based perturbation and permutation. Leveraging the complexity and unpredictability of chaotic systems, the proposed approach perturbs and shuffles model parameters without modifying the standard federated learning protocol on the server side, ensuring secure transmission with minimal computational overhead. Compared with mainstream methods such as differential privacy, the mechanism is lightweight, easy to implement, and causes negligible accuracy degradation. Experimental results on the MNIST and CIFAR-10 datasets show that the maximum accuracy change is only 0.05% and 1.67%, respectively. Meanwhile, the method significantly increases parameter entropy and weakens model inversion attacks, providing a privacy-preserving solution for federated learning with almost no loss in model performance.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.