An Improved Parallel Network Traffic Anomaly Detection Method Based on Bagging and GRU.

In the current large-scale and complex network environment, the types of networks are gradually diversified and the scale is constantly expanding. The network traffic has increased dramatically, and has the characteristics of the high-dimensional multivariable structure, which makes network traffic anomaly detection more and more difficult. Therefore, the paper proposes an improved parallel network traffic anomaly detection method based on Bagging and GRU (PB-GRU). This method uses GRU deep neural network to perform efficient hierarchical feature representation and learn the time-dependent characteristics of network traffic data to achieve more accurate detection. Then use Spark technology to process the training and testing of GRU detector in parallel to improve the overall performance. In order to reduce the individual differences between parallel detectors and improve the generalization error, and Bagging algorithm is used to improve the training process of GRU detector, so that the combined GRU detector has better detection performance. Experimental results show that the proposed method achieves a detection accuracy of 99.6\\(\\%\\), and the error rate is only 0.0036\\(\\%\\). In addition, after parallel processing with Spark, the overall efficiency and scalability have been improved.