Optimized and Efficient Image-Based IoT Malware Detection Method

Abstract

With the widespread use of IoT applications, malware has become a difficult and sophisticated threat. Without robust security measures, a massive volume of confidential and classified data could be exposed to vulnerabilities through which hackers could do various illicit acts. As a result, improved network security mechanisms that can analyse network traffic and detect malicious traffic in real-time are required. In this paper, a novel optimized machine learning image-based IoT malware detection method is proposed using visual representation (i.e., images) of the network traffic. In this method, the ant colony optimizer (ACO)-based feature selection method was proposed to get a minimum number of features while improving the support vector machines (SVMs) classifier’s results (i.e., the malware detection results). Further, the PSO algorithm tuned the SVM parameters of the different kernel functions. Using a public dataset, the experimental results showed that the SVM linear function kernel is the best with an accuracy of 95.56%, recall of 96.43%, precision of 94.12%, and F1_score of 95.26%. Comparing with the literature, it was concluded that bio-inspired techniques, i.e., ACO and PSO, could be used to build an effective and lightweight machine-learning-based malware detection system for the IoT environment.