RPPFL: A Robust and Privacy-Preserving Framework for Federated Learning with Trusted Execution Environments
Abstract
Federated Learning (FL) is vulnerable to privacy breaches and poisoning attacks. We propose a Robust and Privacy-Preserving Framework for Federated Learning (RPPFL) with Trusted Execution Environments (TEEs). RPPFL keeps sensitive layer training separate in a TEE, and its server-side mechanism uses Kullback-Leibler (KL) divergence on intermediate feature distributions to filter clients by scoring unusual behavior. This scoring system, along with a changing median-based threshold, works well to stop both untargeted and backdoor attacks. Tests show that RPPFL stays very accurate and stops advanced attacks, getting more than 60% accuracy (IPM attack) and less than 20% Attack Success Rate (ASR) on CIFAR-10. The work's lies in using KL divergence on intermediate features to create a robust framework for behavioral analysis.
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.