Towards an efficient approach using graph-based evolutionary algorithm for IoT botnet detection

In recent years, a large number of Internet of Things devices are used in life, many of which are vulnerable to attacks from a security perspective. Botnet malware is one of the main threats to IoT devices. Hence detection of IoT botnet is one of the most important challenge for IoT devices. This paper proposes an IoT botnet detection approach based on PSI graph data combine with evolutionary algorithm-based technique. In recent years, a large number of Internet of Things devices are used in life, many of which are vulnerable to attacks from a security perspective. Botnet malware is one of the main threats to IoT devices. Hence detection of IoT botnets is one of the most important challenges for IoT devices. In the paper, a IoT botnet detection approach based on PSI graph analysis by using the evolutionary algorithmbased technique. It applies bacterial evolution algorithm (BEA) in the training process on PSI graph multi-architecture IoT Botnet data to detect IoT Botnet. The PSI graphs were extracted from executable files and transform into vectors to feed into the classical machine learning classifiers. The result of the classifiers is then combine using soft voting method with BEA. The proposed method has achieved good experimental results (i.e., Accuracy at 95.30%, F1 at 96.15%). The approach also achieves a relatively low false-positive rate at 4.59%.