ICAIMT_Real-Time Federated Learning for Autonomous Defense Robots: Enhancing Distributed Threat Detection and Response

Introduction: The intrusion detection system (IDS) analyses huge amounts of data, often containing confidential information, and addresses significant privacy and security problems. Federated learning (FL) has emerged as a promising solution, enabling local model training without raw data exchange, which can protect privacy by increasing detection capacity. The purpose of this research is to develop an FL-based IDS algorithm to solve these challenges and balance privacy, accuracy and calculation efficiency. Methodology: The proposed FL-IDS algorithm appoints FL for model training without compromising data confidentiality. The algorithm is evaluated using multiple datasets including synthetic and real data to assess performance in different scenarios. The most important analysed matrix includes detection accuracy, communication efficiency, convergence and hardware performance. The study addresses the general FL aggregation challenges to ensure strengthening in the distributed environment and includes the privacy mechanisms for privacy guarantee during studies during study training. Results and Discussions: Testing in different datasets confirms the versatility of the algorithm in different environments, and performs better detection rate for both known and new attack patterns. FL-IDS algorithm communication reduces overhead, making it suitable for deployment in the real world. The proposed method improves traditional approaches in data sharing scenarios, and achieves accuracy compared to non-FL methods, proving its viability for privacy-sensitive applications. The BOT-IoT dataset achieved the highest accuracy of 94.1%, followed by ToN-IoT 92.4%, while UNSW-NB15 90.5% had the lowest. The model reached convergence after 18 rounds. Conclusions: The FL-IDS algorithm shows strong performance in the detection of different cyberattacks while effectively protecting data privacy. By taking advantage of FL, it addresses the challenges of detecting major infiltrations, including privacy problems, communication efficiency and compatibility of various attack patterns. Conclusions emphasise their ability as a scalable and secure solution for modern IDS applications, especially in an environment where data privacy is crucial.