Traffic Flow Anomaly Detection and Network Security Protection Strategy Based on Deep Traffic Analysis

Abstract

Abstract: This paper aims to solve the problems of difficulties in precisely detecting traffic flow anomalies and frequent network security risks in intelligent traffic systems. Through deep traffic analysis technology, efficient identification of traffic flow anomalies can be achieved, and network security protection strategies can be dynamically adjusted. This paper combines convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM) models. CNN extracts local spatial features in traffic flow data, and BiLSTM captures temporal dependencies in traffic flow data. The dataset used is Performance Measurement System (PeMS) traffic volume data from California, USA. It collects traffic flow data on working days from February 1 to April 11, 2019, covering traffic flow change data in multiple periods with a sampling interval of 5 minutes. The features extracted by CNN are input into the BiLSTM layer to capture the long-term dynamic changes of traffic flow data, and dynamic weight allocation is performed through the attention layer to help the model focus on the time steps that contribute most to the classification. Network security protection strategies are dynamically adjusted based on traffic flow detection results. The experimental results show that after 70 epochs, the average accuracy of CNN-BiLSTM reaches 99.2%. The classification accuracy of the CNN-BiLSTM model at different time points fluctuates slightly, with a fluctuation of only 0.2%, and its performance is very stable. The average attack success rates of the dynamic network security protection strategy in this paper and the traditional strategy under 10 attack methods are 5.0% and 12.5%, respectively. Combining CNN and BiLSTM for deep traffic analysis and formulating dynamic network security protection strategies can effectively improve the security and stability of intelligent traffic systems and provide new ideas for the intelligent security protection of future traffic systems.