Research On Database User Behavior Pattern Analysis Based On Clustering Analysis And Association Rules

The development of modern information science and network technology puts forward more strict requirements on the security and confidentiality of the database system,the presentdatabase security technology is often difficult to satisfythe demand nowadays.By applying data mining algorithms to analyze the user history behavior information of the database,establish the users' history behavior pattern library,on the basis of this we can effectively detect the misusebehavior of the existing user,thus,achieving the goal of protecting sensitive data in the database.On the basis of previous researches,this thesismainly studies applying the clustering analysis algorithm and association rules algorithmto the database user pattern analysis.The classic K-means clustering algorithm has a need for the user to specify the final cluster number and the initial cluster centers;the classic Apriori algorithm of association rules need to scan the original transaction databasefrequently,and the generated candidate set in the running process of the algorithm is too large,lead to the test of the candidate itemsets need to spend a lot of time.The research work of this thesisis mainly on how to improve the accuracy of database user behavior analysis,while,at the same time improve the shortcomings of the above two algorithms,the specific work of this thesisdescribes as follows:(1)For the problem of manually specified parameters before the K-means algorithm begins,this thesisproposes an adaptive strategy to obtain the final number of clusters and cluster centers,the strategy reduces the impact of the two parameters to the final cluster result,and avoid the final cluster result into local optimum to some extent.(2)For the low efficiency of the Apriori algorithm,this paper proposes a method of using vector array and vertical array to represent the original transaction database,and on the basis of these two data structures reduces the times of scanning the original database and the useless transactions,and the mode matching operation is simplified as well,this all greatly improves the efficiency of the algorithm.(3)This thesisputs forward a database user pattern analysis strategy based on cluster analysis and association rules analysis,this strategy can detect potential abnormalbehavior more effectively comparing to strategies using only one algorithm.Firstly,divide the original database into clusters representing different user role behaviors by using cluster analysis algorithm,use association rule analysis algorithm to analyze each cluster result,dig out the association rules of each user role.Finally,conducting user behavior analysis based on the association rules.The experiment results show that the method proposed in this thesishas better detection results of database internal abnormal behavior.