Hierarchical Obfuscation Malware Detection Method Based on Deep Learning

Abstract:
The new malware hides itself through obfuscation techniques and is able to evade existing detection methods. Artificial intelligence technology is a new direction to solve this problem, however, existing malware datasets have data imbalance problems. Therefore, this paper will propose a deep learning-based hierarchical obfuscation malware detection method, and compare it with the experimental control model and other recently proposed research methods, by standardizing, reconstructing and dividing the CIC-MalMem2022 dataset , build a deep neural network, and build a hierarchical detection model. It is proved by experiments that the classification accuracy of the hierarchical detection model reaches 99%, and it can effectively identify the specific categories of obfuscated malware, which is superior to the existing research methods.