Federated Learning

This paper introduces BART-FL (Backdoor Attack Resilient Technique for Federated Learning), a lightweight defense mechanism designed to detect and filter malicious client updates in federated learning. By combining Principal Component Analysis (PCA), cosine similarity, and K-means clustering with a multi-metric statistical voting system, BART-FL effectively identifies adversarial clients before model aggregation. Experiments on LISA traffic light, CIFAR-10, and CIFAR-100 datasets demonstrate that BART-FL enhances model robustness against backdoor attacks while maintaining high accuracy and computational efficiency.

A hybrid privacy-preserving algorithm combining Pre-Aggregation Similarity Measurement (PA-SM) and Differential Privacy (DP) enhances the security and resilience of Federated Learning (FL) in Intelligent Transportation Systems (ITS) by defending against adversarial attacks while maintaining model performance.

A hybrid privacy-preserving algorithm combining Pre-Aggregation Similarity Measurement (PA-SM) and Differential Privacy (DP) enhances the security and resilience of Federated Learning (FL) in Intelligent Transportation Systems (ITS) by defending against adversarial attacks while maintaining model performance.

This study analyzes secure aggregation and multiparty computation methods in federated learning for autonomous vehicle applications. Using the LISA traffic light dataset, it evaluates their defense against poisoning and inference attacks, highlighting the effectiveness of multiparty computation in preserving privacy and improving model resilience during training and inference.