INTRUSION DETECTION SYSTEM ON FEMTOCELLS OF SMART HOMES: A FEDERATED HYBRID APPROACH

Abstract

The rapid proliferation of Internet of Things (IoT) devices has transformed traditional residences into smart homes, bringing unprecedented convenience but simultaneously exposing users to severe cybersecurity threats. To manage the vast amount of generated data and improve local wireless coverage, femtocell access points have emerged as a critical underlaying communication infrastructure in next-generation networks. However, these femtocells often act as centralized gateways for domestic IoT traffic, creating a vulnerable bottleneck that malicious actors can exploit to gain unauthorized access to private networks. This paper proposes a novel framework for an Intrusion Detection System (IDS) deployed directly at the smart home femtocell level, leveraging a combination of hybrid deep learning and federated learning techniques. By shifting the computational burden away from resource-constrained edge devices to the more capable femtocell access point, the proposed architecture provides robust, real-time threat detection. Furthermore, a federated learning strategy allows multiple smart home femtocells to collaboratively train intrusion detection models with a macrocell without exchanging sensitive, privacy-compromising household data. Through a comprehensive design and hypothetical evaluation plan, this paper demonstrates how combining active learning for alert management with distributed deep learning can significantly enhance the security posture of smart home environments.