Federated Learning-Based Secure Intrusion Detection Framework for Edge-IoT Communication Systems

Abstract

The advent of Edge-Internet of Things (Edge-IoT) communication systems has created a huge cyber security problem with connected devices being heterogeneous, distributed and resource constrained. Centralized intrusion detection systems (IDSs) tend to be data-hungry, impose significant communications overhead, add latency, and have serious privacy concerns. However, due to these limitations, this paper suggests an Edge-IoT Communication Systems Federated Learning-Based Secure Intrusion Detection Framework which facilitates collaborative and secure detection of cyberattacks without sharing raw data from the network. The proposed system includes federated learning systems to enable local ID models training at edge nodes, and distributed edge intelligence to send only encrypted model parameters to a central coordinator. The architecture is built to be used in smart home, smart healthcare, and industrial IoT applications that allow for both the scalable and decentralized cybersecurity operations. It uses a set of deep learning-based IDS models like Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and hybrid autoencoder architectures to identify several cyber threats, including distributed denial of service (DDoS) attacks, botnets, spoofing attacks, ransomware traffic, and abnormal communication behavior. It also includes differential privacy and secure aggregation techniques to ensure privacy and security in local model updates from users, mitigating the risk of inference attacks and unauthorized access. Experimental analysis is conducted using CICIDS2017 dataset and the results show that the proposed federated framework outperforms the centralized IDS and traditional machine learning techniques in terms of high IDS accuracy, low false alarm rates, less communication overhead, and privacy preservation. The findings suggest that the proposed framework offers an efficient, scalable, and lightweight cybersecurity solution for next-generation Edge-IoT communication systems without compromising the information security and adaptive threat intelligence.

Keywords