Research On Detection And Defense Of Malicious Domain Name Based On Deep Learning And SDN Network

With the continuous innovation and development of powerful mobile Internet platforms and modern technologies,people's daily lives are more dependent on the Internet and more intelligent,such as supermarkets' automatic cash register systems,etc.However,as the Internet is convenient for people's daily lives,dangers also arise.The most serious of these dangers is the attack on the network by illegal elements in the Internet.These illegal elements will secretly unknowingly users In the case of obtaining important private information of individuals or companies,especially information about finance,such attacks are extremely harmful.The main transmission medium used by illegal elements is a botnet.With the development of Internet technology,botnets are also different from the previous forms of expression,and the harm is becoming greater and greater.The new technology used in botnets in the new era is domain name generation Algorithm DGA,this algorithm generates a large number of domain names in a very short time and registers,and then binds the IP of the botnet controller,making it difficult for network security personnel to query its real IP.At present,relevant work has been conducted to detect malicious domain names,such as artificially designed feature selection rules and machine learning algorithms.Although these algorithms have achieved some results,there are still some shortcomings,such as manual selection of features has a greater impact on the detection results,and machine learning is more dependent on training data sets.Based on these,this paper proposes a domain name detection method based on a convolutional neural network model.This method can not only automatically learn the deep features of the domain name,but also has a good performance in the accuracy and timeliness of malicious domain name recognition.Finally,the controller of the Software Defined Network is used in the virtual machine system to develop the extended defense system again and conduct simulation experiments,and perform related performance evaluations on the malicious domain name detection model proposed in this paper,as well as the current existing relatively mature feature extraction Comparison of detection methods,the research results show that Software Defined Network(SDN)has great advantages over traditional networks in terms of domain name resolution time and domain name resolution success rate,and the redirection function is turned on in the SDN network.The advantages of traditional networks are more obvious.