Research On Malicious URL Detection Based On LSTM

With the rapid development of communication technology,computer technology and communication technology are more and more popularity.Such as network applications and information sharing are becoming increasingly widespread.The information revolution,machine learning,and neural networks that have exploded around the world are rapidly developing.However,at the same time,the situation of security issues has begun to gradually become more complicated.It can be considered that network information security is a very necessary prerequisite for personal security and social stability.Secure network environment is necessary,condition us to make better use of the network.Therefore,the management and security of the network has become a major issue in scientific research.Using text categorization algorithms and exploiting malicious URLs will greatly enhance web security.The malicious URL recognition technology of machine learning algorithm has wider recognition range and less resources than rule-based recognition technology.Perhaps the malicious URL recognition technology of machine learning algorithms will bring a new breakthrough for the defensive end of web security.In Chapter One we introduces the background,current status,purpose and significance of the research.The current malicious URL attack and recognition status is analyzed.Chapter two,through the Doc2Vec algorithm,LSTM algorithm inspiration,advantages and disadvantages analysis,the URL2Vec-LSTM-Attention neural network architecture is proposed to identify malicious URLs.Then,through experiments,our neural network architecture is compared with the URL2Vec vectorization algorithm and the neural network without Attention mechanism.The results show that our neural network has a good recognition rate.