Cybersecurity Multi-Dimensional Few-Shot Data Generation on Malicious Enhancement

Small amounts of malicious logs can confuse and imbalance a large volume of normal logs, leading to a significant drop in model performance. To address the high heterogeneity and imbalance between network security data, we propose a multidimensional few-shot data augmentation framework based on Generative Adversarial Networks (GANs) for generating high-quality malicious samples to balance data distribution. In this framework, several representative GAN models, including PE-GAN, XOR-GAN, and TriNet-GAN, are designed to enhance the performance of deep learning models in network threat detection. Large-scale adversarial generation experiments are conducted on the original imbalanced dataset and security event logs using AC-GAN and Seq-GAN, effectively solving the imbalance and few-shot issues. The experimental results are based on the publicly available NIMS (Network Information Management and Security Group) and KDD99 datasets. The results demonstrate that the proposed method performs well in data augmentation for few-shot, imbalanced, and multidimensional complex data.