Anomaly detection framework for Internet of things traffic using vector convolutional deep learning approach in fog environment

Abstract

The proliferation of Internet of things (IoT) devices has lured hackers to launch attacks. Therefore, anomalies in IoT traffic must be detected to mitigate these attacks and protect services rendered by smart devices. The lacuna in the existing anomaly detection techniques is the nonscalable nature of anomaly detection systems, resulting in the mishandling of large-scale data generated from IoT devices. The issue of scalability is addressed and an anomaly detection framework in a fog environment is proposed herein using vector convolutional deep learning (VCDL) approach. The anomaly detection system could be scalable if the traffic can be distributed to the nodes in the fog layer for processing. This is effectively captured in the VCDL approach in which the training of IoT traffic is distributed and computations are performed in the fog nodes. The parameters required for training are shared by the master node in the fog layer. Further, the proposed anomaly detection algorithm classifies IoT traffic as either normal or attack and then passes it to the cloud for attack mitigation. Experiments were conducted on UNSW’s Bot-IoT dataset and the results indicate that the proposed distributed deep learning approach can efficiently handle scalable data compared with the existing centralized deep learning approaches. Experimental results show that the proposed approach is significantly better in terms of accuracy, precision, and recall compared with the state-of-the-art anomaly detection systems.