Research And Design Of Malicious URL Detection Algorithm Based On Deep Learning

In recent years,with the continuous evolution of web attacks,many network applications are suffering from various forms of security threats and cyber attacks.More and more security vulnerabilities are frequent,and traditional network security detection is gradually exposing defects.This urgently requires a new type of method to accurately detect network attacks,so the use of deep learning for network attack detection has become one of the hot research areas.This paper firstly studies the existing malicious URL detection technology,applies the deep learning method to the URL attack behavior recognition,and designs a keyword-based neural network model for malicious URL detection.The method utilizes the data set obtained from the security vendor,combines the word embedding technology and the malicious keyword in the URL,uses the convolutional neural network to extract the feature of the URL,and uses the GRU to perform feature capture in the time dimension.Experiments show that the proposed detection model can obtain high-precision multi-classification results.Secondly,this paper also proposes a model updating algorithm based on pruning compression,which can update the parameters and structure of the network model at the same time.The CNN pruning and SVD are used to remove the redundant connection of the model to dynamically update the model structure.Compared with the traditional model updating algorithm,the generalization ability of the neural network is improved,and the classification effect is better.Finally,in order to use the detection model for actual network security protection,this paper designs and implements a set of visual detection system for malicious URLs.In order to verify the performance of the detection system,this paper builds a simulation environment to test and analyze the system.The results show that the accuracy of the detection system is above 98%,and the detection results can be displayed more intuitively.