Enhancing Intrusion Detection in IoT-Based Vulnerable Environments Using Federated Learning
Nallagatla Raghavendra Sai, G. Sai Chaitanya Kumar, Dasari Lokesh Sai Kumar, Surapaneni Phani Praveen, Thulasi Bikku · 2024
With the rapid proliferation of Internet of Things (IoT) devices, vulnerable environments face increasingly sophisticated cyber threats, making intrusion detection a critical security concern. Traditional centralized intrusion detection systems struggle to address the scale and diversity of IoT networks while protecting sensitive data. In this research, we explore the application of federated learning, a decentralized machine learning paradigm, to enhance intrusion detection in IoT-based vulnerable environments. Our research focuses on designing a federated learning framework tailored to IoT settings, considering communication constraints, data privacy, and resource limitations. We propose a novel model architecture that enables collaborative learning across distributed IoT devices while preserving individual data privacy. We deploy the federated learning model on a realistic IoT testbed featuring various devices and sensors to validate our approach. Through extensive experiments, we evaluate the performance of our federated learning-based intrusion detection system against traditional centralized models. Our results demonstrate the effectiveness of federated learning in improving intrusion detection accuracy, especially in scenarios with diverse and dynamic IoT device populations. Furthermore, we assess the model&s;s robustness against adversarial attacks and its adaptability to changing IoT environments. However, we acknowledge the security and privacy challenges posed by federated learning, particularly in resource-constrained environments. Therefore, we discuss privacy-preserving techniques and mitigation strategies to protect sensitive data during the collaborative learning process. In conclusion, our research presents a novel approach to enhance intrusion detection in IoT-based vulnerable environments through federated learning. The study contributes to cybersecurity by addressing centralized systems&s; limitations and paving the way for scalable and privacy-aware intrusion detection in the IoT era.