Research And Implementation Of Malicious Website Detection Model Based On Content In Mobile Environment

With the popularity of mobile Internet,mobile payment and mobile shopping have become daily.At the same time,Internet crimes such as phishing,malware and so on are becoming more and more serious.Criminals posing as banks,electricity providers,social networking sites to send fraudulent information to induce users to log on,steal user information,so that the vast numbers of users and financial institutions suffered property and economic losses.In recent years,the security of personal information has attracted more and more attention.It is very important to identify the significance of network risk related to malicious website accurately and effectively.At present,both domestic and foreign research on phishing sites have achievements,but there are shortcomings.The detection method based on black-and-white list is less timeliness,honeypot based is slow response,the accuracy and generalization of the machine learning algorithm based on content features needs to be improved.In recent years,deep learning has been applied to various fields and has achieved great success.In view of the above,this paper analyzes the existing method of detection of malicious website,presents a malicious website detection method based on Deep Belief Networks.What's more,combined with various detection methods,a hierarchical malicious web site analysis system is designed and implemented.The main innovative work of this paper is summarized as follows:1)Aiming at the time-consuming problem of image similarity processing in the malicious website analysis model based on similarity detection,this paper optimizes the implementation process of the perceptual hash algorithm,thus the processing speed of image similarity detection is improved.And the local sensitive hash is introduced into the domain similarity detection,which improves the accuracy and performance of the domain name similarity detection.2)Aiming at the decline of model recognition accuracy caused by disequilibrium of sample data in content based malicious website detection model,this paper uses the Borderline-Smote up-sampling method to solve the problem and applies the DBN model to the malicious website detection.The experiment shows that the Borderline-Smote DBN algorithm effectively alleviates the problem of reducing the accuracy of the model recognition in the data imbalance.And the algorithm has higher accuracy and lower error rate than the popular machine learning algorithm.3)In this paper,we design and implement a hierarchical malicious website analysis system,combining black-and-white list detection,similarity detection and deep learning model,filter layer by layer,cooperate with each other,use the light detection strategy to enhance the response speed of the system,and use the deep learning model to improve the accuracy of recognition.