TCB Intrusion Detection Method Based on Data Enhancement

Within the online sphere, intrusion detection technology plays an increasingly vital role in safeguarding our information security. The current research on intrusion detection mainly focuses on two aspects: machine learning and deep learning, but it cannot be ignored that data balance also affects the efficiency of intrusion detection. The vast majority of current datasets are imbalanced, and during the training process, the model ignores a few class samples, which affects the efficiency of intrusion detection. Therefore, in response to this issue, this article introduces a novel intrusion detection model that integrates Transformer and CNN-BiLSTM, leveraging WGAN-enhanced data. This model first uses Transformer for global capture, establishes global connections, then uses CNN-BiLTM to capture spatial and temporal features, and finally uses Softmax classifier for classification. The dataset used in this article is NSL-KDD. The experimental results show that after data balance processing, the accuracy of two classification tasks of TCB model reaches 92.19%, which is at least 2.24% higher than other models, and the accuracy of multi classification tasks reaches 82.94%, which is at least 1.03% higher than other models.