Research On Network Traffic Anomaly Detection Based On Deep Learning

With the development of science and technology,the Internet plays an important role in people's life and economic development.But at the same time,network attacks are also showing a growing trend year by year,posing a great threat to network security and network order.Network traffic anomaly detection methods based on mathematical statistics and data mining generally have high false positive rate and low accuracy.The performance of traditional machine learning algorithms is highly dependent on features,so how to construct a universal network traffic feature set is an unsolved problem.With the rise of deep learning,it brings new opportunities for the research and development of network traffic anomaly detection.This paper conducts research on the above issues,and its main contributions and innovations are as follows:1.This paper designs a network traffic binary classification model based on deep learning Firstly,the discrete features are transformed into low-dimensional and dense feature vectors by Embedding technology,which effectively preserves the semantic information between the discrete features.The model combines the attention mechanism with the traditional fully connected neural network to enhance the representation ability of the model.The model is a lightweight model,which takes into account the accuracy of the model and the delay of the system.2.This paper designs a network traffic multi-classification mode based on deep learning The highlight of this method is the application of Deep FM framework to anomaly detection tasks.Deep FM takes into account the advantages of both deep learning and traditional machine learning.The FM part extracts loworder features with memory ability,while the neural network part extracts high-order features with generalization ability.By integrating these two features,the accuracy of model classification is effectively improved.The application of FM algorithm effectively solves the dependence of machine learning algorithm on artificial features and enables the model to automatically construct cross features.In the neural network part,a-Resnext module is proposed,which combines grouping convolutional network with self-attention mechanism.Grouping convolution can extract rich features,and attention mechanism can reduce the importance of redundant features to some extent,and improve the representation and generalization ability of the model.3.This paper proposes a two-stage network traffic anomaly detection system based on deep learning In the first stage,the network traffic binary classification model based on deep learning can quickly divide the traffic into normal traffic and abnormal traffic.In the second stage,the network traffic multi-classification model based on deep learning is used to further subdivide abnormal traffic.The system combines accuracy and system delay,and can label network traffic to form a data loop.Experiments show that the two-stage detection method can detect most of the flow quickly and accurately,and meet the needs of industrialization.