Optimal feature selection for malware detection in cyber physical systems using graph convolutional network

Abstract

Due to the fast growth of Internet of Things (IoT) technology, the detection and analysis of malware have become a challenge for industrial applications of Cyber-Physical System (CPS), which delivers a variety of services based on the IoT paradigm. The malicious programme presents a security concern to CPSs because to their reliance on the internet, ICT services and commodities, and the internet itself. The purpose of this research is to build a Snake optimizer-based feature selection with optimum graph convolutional network for malware detection (SOFS-OGCNMD) for the CPS environment. The primary purpose of the SOFS-OGCNMD model is the identification and categorization of harmful software in a CPS environment. To recognize and categorize malware assaults, the SOFS-OGCNMD model use the flower pollination algorithm (FPA) in combination with the GCN method. The design of the FPA permits the selection of the most appropriate parameters for the GCN method, hence enhancing the detection performance. The experimental validation of the SOFS-OGCNMD system is studied using benchmark datasets, and comparative analysis indicates that the system's average accuracy is 98.28%, its average precision is 98.65%, its recall is 98.53%, and its F1-Score is 98.47. The findings indicate that the SOFS-OGCNMD technique outperform more recent models.