Research and implementation of Android malware detection algorithm based on Graph Convolutional Networks
Yue Wang, Hailati Kezierbieke, Qinglin Chen · 2024
Android malware detection is one of the research hotspots in the field of malicious software. Traditional rule-based and feature-based methods face challenges in handling large-scale Android malware detection. This paper proposes an Android malware detection model based on graph convolutional network algorithms. Additionally, an ensemble learning method is utilized, selecting three different graph convolutional layer algorithms as base classifiers. The base classifiers are trained and tested to obtain accuracy, precision, recall, and F1 score as evaluation metrics, and the weights of the classifiers are calculated using these metrics. This approach avoids the potential distortion in classification results from a single graph convolutional network algorithm, thereby enhancing its classification performance. This paper also introduces a ReliefF feature selection algorithm based on mutual information optimization. By utilizing mutual information from information theory to calculate the correlation between features and improving the computation of feature weights, it addresses the issue of handling redundant features in the traditional ReliefF algorithm to a certain extent.