Rapid and efficient Android malicious software detection method based on multiple features

The invention discloses a multi-feature-based rapid and efficient Android malicious software detection method, relates to the field of smart phone information security, and solves the problems that an existing machine learning algorithm using a single feature cannot give full play to the data processing capability and is poor in detection effect, and although a multi-feature machine learning algorithm and the detection effect is improved, a large amount of time and high-quality hardware are required to meet the detection conditions. According to the invention, the CatBoost algorithm is applied to the field of Android malicious software detection and classification, permission and Dalvik operation code features are combined, the features are segmented through an N-Gram method, a feature dimension reduction method is designed for processing the features, Android application information can be represented more comprehensively, and therefore a feature model can be established more accurately. According to the det