Multi-class Malware Detection via Deep Graph Convolutional Networks Using TF-IDF-Based Attributed Call Graphs

doi:10.1007/978-981-99-8024-6_15

2024-01-11

PRE

Irshad Khan

Youngwoo Kwon *

DOI：10.1007/978-981-99-8024-6_15

原文链接

原文求助

摘要

En 中文

The proliferation of malware in the Android ecosystem poses significant security risks and financial losses for enterprises and developers. Malware constantly evolves, exhibiting dynamic behavior and complexity, thus making it challenging to develop robust defense mechanisms. Traditional methods, such as signature-based and battery-monitoring approaches, struggle to detect emerging malware variants effectively. Recent advancements in deep learning have shown promising results in Android malware detection. However, most existing approaches focus on binary classification and need more insights into the model's generality across different types of malware. This study presents a novel approach to address Android malware detection by integrating TF-IDF (Term Frequency-Inverse Document Frequency) features into the call graph structure. By attributing each node in the call graph with TF-IDF-based feature vectors extracted from the opcode sequences of each method using an opcode list, we present a more thorough representation that encapsulates the complex traits of the malware samples. We employ state-of-the-art graph-based deep learning models to classify malware families, including Graph Convolutional Networks (GCN), SAGEConv, Graph Attention Networks (GAT), and Graph Isomorphism Networks (GIN). By incorporating high-level structural information from the call graphs and TF-IDF-based raw features, our approach aims to enhance the accuracy and generality of the malware detection models. We identify an optimal model for the Android malware family classification task through extensive evaluation and comparison of the above-mentioned models. The findings of this study contribute to advancing the field of Android malware detection and provide insights into the effectiveness of graph-based deep learning models for combating evolving malware threats.

Keyword:

Malware

TF-IDF

call graph

graph convolutional model