Traffic Anomaly Detection Model Integrating Conditional Generative Adversarial Network and WaveNet

To address the issues of low detection rates for minority classes and data imbalance in traffic anomaly detection, this paper proposes a new traffic anomaly detection model. Specifically, a traffic anomaly detection model that integrates Conditional Generative Adversarial Networks (CGAN) and WaveNet-BiL STM is presented. The process begins by converting raw traffic data into numerical features, followed by the use of the Interquartile Range (IQR) method to remove outliers and normalize the data. Next, CGAN is employed to balance the dataset. The preprocessed data is then fed into WaveNet and Bidirectional Long Short-Term Memory (BiLSTM) networks for feature extraction. Concurrently, the Sparrow Search Algorithm (SSA) is utilized for hyperparameter optimization, selecting the most suitable hyperparameters for the model to minimize the impact of parameter selection on model accuracy. This approach enhances the detection accuracy for minority attack classes. Experimental results demonstrate that the proposed model achieves an accuracy of 89.98% and an F1 score of 90.92% on the NSL-KDD dataset.