FSAC-AFedAvg: Federated Soft Actor-Critic Reinforcement Learning-Based Intrusion Detection System with Adaptive Federated Averaging

Recent advancements in Internet of Things (IoT) and cloud computing technologies have made intrusion detection and prevention in enterprise networks increasingly complex. Traditional intrusion detection systems (IDS) often rely on centralized data collection, raising privacy concerns and limiting adaptability to evolving attack patterns. To address these issues, we propose FSAC-AFedAvg, a Federated Soft Actor-Critic (FSAC) reinforcement learning–based IDS with adaptive federated averaging, designed for distributed and privacy-preserving intrusion detection. Our approach integrates federated learning with adaptive client weighting to improve global model convergence under heterogeneous data, while SAC enhances detection performance by optimizing decision-making based on dynamic network conditions. Evaluations on CSE-CIC-IDS2018, MQTTset, and InSDN datasets demonstrate that FSAC-AFedAvg achieves high detection accuracy, low false-positive rates, and strong scalability, outperforming existing IDS solutions in distributed environments.