Research On Detection Of Malicious Software

In recent years, with the rapid development of the Internet, the Internet has become an indispensable part of society. However, with the increasing variety of malware and the application of encryption methods, the security of Internet network is constantly threatened. How to detect and identify malicious software without affecting the normal operation of user hosts and violating user privacy by monitoring a small number of non-sensitive features while software is running dynamically, so as to protect user host information, has become an imminent issue in the field of network security. In this work, a new feature extraction method is developed and proved to be effective. This paper presents a characterization method to extract malware features from three aspects: derived features, vector space features of API and context features of API. XGBoos, LGBM, Improved TextCNN models are trained to predict test sets. Finally, these models are combined with Stacking model to output the final results.