Malicious Network Traffic Detection in IoT Environments Using A Multi-level Neural Network

Abstract

The Internet of Things (IoT) is a system that connects physical computing devices, sensors, software, and other technologies, and data can be collected, transferred, and exchanged with other devices over the network without requiring human interactions. One challenge that the development of IoT is facing is the existence of malicious botnet attacks. Recently, research on botnet traffic detection has become popular. However, most state-of-the-art detection techniques focus on one specific type of device in IoT or one particular botnet attack type. Therefore, we propose a neural network-based algorithm, 2-FFNN, which can detect malicious traffic in the IoT environment and be deployed generally without restricting device or attack types. The proposed model consists of two levels of the Feed Forward Neural Network framework to identify some hard-to-detect botnet attacks. Experimental analysis has shown that the 2-FFNN outperforms the baseline FFNN and some state-of-the-art methods based on the detection accuracy and ROC score.