Android Malware Detection Based on Logistic Regression and XGBoost
Suhuan Li, Huang Xiaojun · 2019
The proliferation of Android malicious applications dangerously injure users' information, property, and privacy. Aiming at the problem that the characteristics of malware dynamic analysis and detection aren't excellent, and the detection efficiency and classifier performance are insufficient, this paper proposes a multi-dimensional feature fusion malicious application detection method based on Logistic Regression and XGBoost. The method provided non-invasively extract the framework layer Application Programming Interface(API) call information of the Android application, apply the Logistic Regression to train the N-gram modeled API call sequence, fuse the obtained probability feature with the basic statistical feature, and input it into the XGBoost for Android malicious application detection. The experimental results show that the method effectively improves the accuracy of Android malicious application detection and decreases the time expense of it.