Research on Network Traffic Anomaly Detection Method Based on Deep Learning

Abstract

Network attacks are a serious problem in today's information society. Intrusion detection systems are an important guarantee for network security. The core of IDS is the intrusion detection algorithm. Under the current new situation of network security protection, traditional signature-based misuse detection algorithms cannot solve the problem of rapidly changing forms of network attacks. Machine learning and data mining methods have been widely used in attack detection in the network environment to improve the detection rate. Based on the in-depth analysis and research of deep learning models and intrusion detection methods, this paper proposes a convolutional neural network abnormal traffic detection model based on dynamic adaptive pooling for the pooling layer can only use one fixed method for down sampling, and proposed an abnormal traffic detection framework and a dynamic adaptive CNN model. Experimental analysis shows that the method proposed in this paper shows better detection accuracy and loss value in traffic detection.