Deep Learning-based Worm Detection Method for Polymorphic Networks

The maintenance of cybersecurity has become increasingly vital with the advancement of the information age, particularly in the detection of polymorphic network worms. Due to their ability to mutate, these worms can evade the capture of traditional detection techniques, for which the application of deep learning provides a new perspective. This paper delves into the exploration of deep learning, especially Convolutional Neural Networks (CNN), in the field of polymorphic network worm detection. By studying deep learning algorithms and understanding how they operate in complex pattern recognition tasks, this study constructs a detection framework based on CNN for the identification and disruption of worm propagation. In addressing network traffic data, a unique method of effective payload matrix processing is proposed to meet the data input requirements of CNN. This method can extract deep-seated features hidden in traffic data, enabling the effective identification of polymorphic worm attacks. Through the establishment of an experimental environment and the validation of the proposed method using standard datasets, the experimental results confirm the accuracy and practicality of this approach in polymorphic worm detection. This work significantly enhances the capability of the network security defense system and lays a foundation for defending against future worm attacks.