Two-path Android Malware Detection Based on N-gram Feature Weighting

Abstract

In recent years, with the full popularity of Android system and applications, the types and number of Android malicious applications also show explosive growth, and more efficient detection technology is urgently needed to identify malicious software. In view of the current research on N-gram features is relatively single, in order to make more comprehensive use of N-gram features and explore the potential relationship between features and attributes of applications, this paper proposes a two-path Android malware detection model based on N-gram feature weighting, and achieves N-gram feature extraction in two different ways by setting an application file threshold. Finally, Neural network is used to classify the fused features. Testing results of 1205 malicious samples and 1084 benign samples shows that the detection accuracy of the model was up to 99.2%. At the same time, this experiment further verify the effectiveness of relevant improvements, and the results show that compared with traditional machine learning algorithms, this model has higher adaptability and accuracy.