Graph Convolutional Network for Detection
Tony Thomas, Roopak Surendran, Teenu S. John, Mamoun Alazab · 2022
This chapter discusses the application Graph Convolutional Network (GCN) in Android malware detection and illustrates with the detection of obfuscated Android malware from system call graphs. It gives an introduction of GCN and its applications in many real-world problems. The chapter explains GCN-based malware detection. GCN models are neural network models that can learn the graph structure and can aggregate the node information in a convolutional fashion. There are two types of GCN: spatial-based GCN and spectral based GCN. Malware developers can employ a variety of mechanisms to circumvent its detection by generating malware variants that mimic legitimate applications. Android malware continue to emerge day by day and hence it is challenging to detect the malware in an effective and scalable manner. The chapter explores whether GCN can detect the evolving Android malware. It presents the ideas to help cyber security researchers and developers to unveil the potential of GCN in developing automated cyber security solutions.