Privacy - Preserving Anomaly Detection Using Federated Learning and Explainable AI

Abstract

Anomaly detection is crucial for identifying security threats and system failures. Traditional methods often require centralized data collection and raising privacy concerns. This paper proposes an idea of privacy-preserving anomaly detection system using Federated Learning (FL), Explainable AI (XAI) and Generative Adversarial Networks (GAN). Federated Learning provides with decentralized training while preserving the data privacy and Explainable AI enhances model transparency, helping in decision making. By utilizing deep autoencoders for anomaly detection and SHAP/LIME for explainability, it ensures secure and interpretable anomaly detection across distributed environments. The proposed model is trained and validated using realworld datasets, demonstrating effectiveness in identifying anomalies while minimizing privacy risks