A Method for Generating Characteristic Samples of Imbalanced User Attack Behavior Based on DD-DCGAN

Due to the imbalance of user attack behavior characteristic samples, user attack behavior detection models are faced with problems such as low detection accuracy and poor performance when dealing with minority-class samples. To address the above issues, this paper proposes a method for generating characteristic samples of imbalanced user attack behavior. Firstly, collect the traffic data of user attack behaviors in open-source software systems, manually label the attack types, calculate the data imbalance degree, and then fuse the generated RGB feature maps and information entropy single-channel grayscale maps to generate minority-class RGBA feature maps and majority-class RGBA feature maps.The improved dual-generator and dual-discriminator deep convolutional generative adversarial network (DD-DCGAN) model is then employed, with the loss functions for the dual generators and discriminators optimized. By continuously training the DD-DCGAN model, a Nash equilibrium state is achieved, generating minority class RGBA feature maps that provide high-quality sample data for user attack behavior detection and classification. The model comparison results show that this method performs well in generating user attack behavior feature samples, demonstrating greater robustness and reliability.