Research On Key Technologies Of Website Defacements Detection Method

Due to the characteristics of Web services,website defacements can be widely spread within a short period of time and among various groups of people.The proliferation effect is extremely rapid and the scope is extremely wide.The impact is extremely bad.It has become a hot topic in the current field of cyber security research to detect the website defacements effectively,especially for the emerging websites to be detected in time.Defaced websites have many problems such as the ubiquitous phenomenon of website trojan,poorly designed,simple structure,single content,and the difference between the visual effects of defaced webpages and normal webpages,this paper studies the problem of website defacements detection from the angle of the behavior of website trojan,text and structure of webpage,and computer visual.Firstly,this paper uses the distributed data acquisition technology based on Scrapy-redis to periodically crawl website data and update storage.According to the principle analysis of website trojan behavior,the paper proposes a website trojan detecting method based on rule matching and builds a website trojan knowledge base.The detection is also constantly enriching the website trojan knowledge base.Secondly,this paper proposes a website defacements detection algorithm based on webpage text and structure.The algorithm extracts the text and structural features of webpages.In order to improve the accuracy and stability of detection,the feature selection algorithm based on SVM-RFE is used to gradually eliminate the redundant features to obtain the optimal feature subsets.And the SVM classification algorithm is used to classify webpages to achieve the website defacements detection.Experimental results show that the website defacements detection algorithm using SVM-RFE feature selection algorithm combined with SVM can achieve 96% accuracy.Then,this paper proposes website defacements detection algorithm based on webpage screenshots.The algorithm uses webpage snapshot generation technology,webpage screenshot window extraction technology and webpage subgraph normalization technology to perform data preprocessing.In order to reduce the workload of manually extracting features,the paper adopts the stacked autoencoders to automatically learn the high-dimensional features of the screenshot,and the convolution neural network classification algorithm to improve the classification effect of webpages.At the same time,a strategy for fine-tuning neural network is adopted to effectively avoid the "concept drift" phenomenon.Experimental results show that the accuracy and recall rate of the classification algorithm based on stacked autoencoders and convolutional neural network in website defacements detecting have reached 90%,which shows the stability of the detection algorithm.Finally,on the basis of researches above,we designed and implemented a website defacements detection prototype system.The system realizes the data collection,website trojan detection and multi-angle website defacements detection,and then proposes an integrated decision strategy to improve the accuracy of detection.The system test results show that the system has better performance in website defacements detection.