Behavioral Pattern Mining for User Identity and Access Control A Cluster based Ensemble Model

Web usage mining extracts user's behavior patterns from the internet. Behavior of web users services are monitored and controlled to authenticate the user identity and access. Several data mining techniques are presented to analyze the web user behavioral patterns. But complex activity pattern discovery is not performed to maintain the decision making. To improve the complex user behavior pattern mining, Ensemble of Fuzzy K-Means with Logit Boost Clustering (EFK-LBC) technique is developed. EFK-LBC technique extracts the web user behavioral patterns from web logs. First, preprocessing is exploited to clean the unwanted data and select consistent web patterns from the original web log files. Next, fuzzy k means clustering technique is employed as a base learner to group the frequent web user behavioral patterns based on the objective function. To improve clustering performance, Logit Boost clustering technique is designed to make strong cluster by combining the several base learners. Experimental evaluation of proposed EFK-LBC technique and existing methods are carried out with the web server log files. The results reported that the proposed EFK-LBC technique obtains high clustering accuracy of user identity with minimum time and space complexity. Based on the observations, EFK-LBC technique is more efficient than the existing methods.