Research On Web Fingerprint Identification

Web fingerprint is the characteristic information left by the developers during the development of website components,which can indicate the type and version of the components.By identifying these characteristics information,the components used by the website can be identified.While the component brings convenience to the website construction,it also brings security risks.Now,many hacker attacks target components,and a bug of the component could lead to a large number of sites being attacked.Therefore,it is necessary to fully explore the website component resources,when there is a component vulnerability,it can timely respond to early warning,reduce the scope of component vulnerability impact,which is of great significance to maintain the security of cyberspace.At present,the mainstream web fingerprint identification technology is based on feature matching.The accuracy of web fingerprint identification is limited by the website component fingerprint database.And the discovery of component fingerprints mainly depends on manual annotation.Secondly,the feature information of website component is easy to be modified by developers,especially for the web server.It is easy to modify or disguise the Banner information,which often leads to identification failure.This thesis mainly studies on methods of website component fingerprint collection and web server identification technology.The main work of the thesis is as follows:(1)Aiming at the problem of fingerprint collection of website components,this thesis studies the structure and distribution of existing component fingerprints,and a method for discovering component fingerprints is proposed.This method is realized by comparing the digital summary of static files of components.Finally,adding 341 website components and754 component fingerprints.Secondly,compared with the traditional website component fingerprint,the associated component fingerprint is introduced.Components are identified by identifying the components associated with them.This thesis uses two methods to discover related component fingerprints: a dictionary-based method for extracting related components,and a method based on the characteristics of component source files.Finally,the associated components information for 685 web components was obtained.(2)Aiming at the problem of web server identification,this thesis turns web server identification into a multivariate classification problem based on the different web servers' handling of the same HTTP request.Based on the machine learning classification method,the results of multiple classifiers are fused to identify the web server.Compared with a single classifier,the accuracy of the web server identification is improved.