Fileless Memory Trojan Detection Method Based on Deep Learning

This paper proposes a fileless memory trojan detection method based on deep learning, which is specifically aimed at attack methods represented by advanced fileless memory trojans such as Ice Scorpion. First, this paper constructs a database covering common memory trojan signature features. Then, a dynamic detection model is designed, which learns abnormal behavior patterns in memory through deep neural networks, so that it can detect and identify potential memory trojans in real time without relying on file features. To verify the effectiveness of the model, this paper uses multiple Ice Scorpion memory trojan samples for testing in simulation experiments. Experimental results show that the detection method proposed in this paper has high accuracy and recall rate, with a detection accuracy of 98.5% and a recall rate of 97.2%, which is significantly better than traditional signature-based static detection methods. In addition, the algorithm also has good performance in processing efficiency, and can achieve real-time detection without significantly increasing system resource consumption.