Binary and Multiclass Classification of Attacks in Edge IIoT Networks

The Internet of Things (IoT) refers to a smart network of digitally connected physical devices and systems, with applications ranging from predictive maintenance, optimization, enhanced efficiency in manufacturing and critical infrastructure. Over the years the applications of IoT have been extended from domestic to Industrial Internet of Things (IIoT). The increased integration of IoT in real-world applications exposes the networks to cyber threats, e.g., Man in the Middle and ransomware attacks. Such attacks on IoT networks impact their operational integrity, data confidentiality, and overall system reliability, especially in IIoT networks. Therefore, the timely and accurate detection of cyber attacks can save these networks from unauthorized access, malware propagation, and denial of service attacks. In this paper, we present a cyber attack detection technique using Edge-IIoTset dataset for IIoT applications. The proposed detection approach is based on binary and multi-class classification of cyber attacks in IIoT networks, leveraging machine learning (i.e., logistic regression, decision tree) and deep learning (i.e., recurrent and convolutional neural networks) algorithms. The proposed classification techniques showcase the effectiveness of the applied algorithms achieving an average accuracy of 90 %.