MQTT Anomalous Behavior Detection in IP Sensor Networks Through Convolutional Neural Networks and Traffic to Image Encoding

The rapid growth of Internet of Things (IoT) devices in sectors like healthcare, manufacturing, and smart cities has resulted in a substantial increase in data volume and complexity. This requires robust anomaly detection systems to identify critical issues such as system failures, security breaches, external attacks, and inefficiencies. However, traditional anomaly detection methods often struggle with the high-dimensional and dynamic nature of IoT data. In this paper we propose a new and unconventional approach for anomaly detection, cyber-attacks in particular, in IP sensor networks, based on encoding IP packets into image and exploiting the huge and well-known classification strength of Convolutional Neural Networks for malicious behavior recognition. Simulation results show the optimality of the proposed machine learning approach, outperforming the existing tools in terms of accuracy, time and complexity.