Research on Recognition of Android Counterfeit Application Based on Siamese Network
Kun Li, Yanyan Liu, Xuchen Wang, Guopeng Li, Ziyue Ma · 2023
In recent years, with the popularity of mobile phones and mobile Internet, the download volume and application rate of third-party applications for smart phones, namely APP, have increased rapidly, and criminals have taken advantage of users' trust in well-known apps to produce similar interfaces and functions, induce users to download and install, steal users' personal information, property or spread malware, not only causing losses to users, but also disrupting normal market order. In order to identify counterfeit APP more quickly and effectively, this paper proposes a counterfeit APP recognition model based on twin network. The model first decompiles and extracts the name, package name, icon, and signature information of the APP, then filters out suspected counterfeit applications by calculating the editing distance of the name, and finally calculates the icon similarity based on the twin network model to determine whether the application is counterfeit. In this paper, datasets containing multiple types of phishing applications are used for experiments, and the effects of VGG16, ResNet50, and ViT algorithms on the recognition results are compared as the basic feature extraction networks. The results show that the accuracy of the ViT-based twin network architecture reaches 85.12%, which can effectively identify counterfeit applications.