Research And Implementation Of Sliding Event Window Based Malicious Download Detection System For Library Resource

The malicious download of library resource has occurred frequently in universities.The detected malicious IP/IP segments' accesses will be forbidden by the resource providers,which will seriously affect normal users.Therefore,the research on malicious download detection of library resources is of important research significance.Most current detections are through the detection of abnormal traffic,excessive download frequency or abnormal access request content.They have problems such as fuzzy quantification and high false positive rate,and are not effective for low-speed random time interval crawlers.Most malicious download detection methods are IP-based,which cannot identify malicious users when Network Address Translation or IP address spoofing is used.To solve problems above,this thesis proposes a malicious download detection algorithm based on sliding event window for library resource.Sliding event window realizes time randomized detection.The topic relevance of library resources is used as the basis for malicious download detection,which is achieved by topic clustering algorithm.Based on the proposed detection algorithm,this study designs and implements a library resource malicious download detection system.The system obtains user download request information and related titles or s through the library resource theme data collection module.Then the data is associated with the user account in[IP?Account]mapping module.The user behavior data processing module extracts the topic sentence from abstracts,unifies data format and builds sliding event windows according to user accounts.The malicious download detection module clusters the download events in the window through topic relevance,so as to detect malicious download.This system has been tested in a university library.The result shows that on the basis of the traditional feature-based detection,this system can well detect malicious download including low-speed random time interval crawlers with topic relevance analysis.