Edge Multi-agent Intrusion Detection System Architecture for IoT Devices with Cloud Continuum

The Industry 4.0 has brought significant changes in production processes and business models worldwide. Advanced technologies, e.g., Collaborative Robotics, Artificial Intelligence, Cloud Computing, and Internet of Things (IoT) are playing a crucial role in improving efficiency and productivity. However, the adoption of these technologies, particularly IoT, introduces security vulnerabilities and potential attacks due to inadequate security measures. This paper addresses the need for dedicated cybersecurity mechanisms and secure device design in IoT networks, particularly emphasizing the challenges faced in implementing Intrusion Detection Systems (IDS) on resource-constrained IoT edge devices, limiting the use of traditional machine learning based detection methods. Moreover, the limited computational resources of IoT devices require lightweight techniques that have low power requirements but can accurately detect anomalies in the network. To tackle these challenges, a novel multi-agent based architecture is proposed, considering the distribution of nodes along the edge-cloud continuum, and enabling the collaboration among different processes to detect anomalies during attacks. The proposed architecture is evaluated at the edge level using the CICIoT2023 dataset. The results demonstrate the feasibility of using multi-agent systems for a collaborative detection of IoT attacks, contributing to enhance the security of IoT-based systems against cyber threats in Industry 4.0 environments by leveraging lightweight techniques.