Research On Database Intrusion Detection Based On Random Forest

The security of database has been attracting much attention since it is a very vulnerable link in the information system.As the most important part of security mechanism,intrusion detection is a frequently used security technology.The classification algorithm plays a crucial role in the intrusion detection system whose essence is classification.Random Forest has the advantage of integrated learning theory,efficient data processing ability and simple parameter setting.This paper studies the application of random forest algorithm in the database intrusion detection.This paper introduced the related content of database security and intrusion detection,researched the methods of common attack on database and the main techniques of database intrusion detection,analyzed the advantages and disadvantages of the application of Decision Tree,Gradient Boosting Decision Tree and Random Forest in the intrusion detection.In order to solve the problem of insufficient accuracy of random forest algorithm meta-classifier,this paper presents a random forest database intrusion detection algorithm with the decision tree of gradient enhancement as the meta-classifier.It increased the classification accuracy,improved the resampling of the original data set in the algorithm and decreased correlation of noise data.The random voting mechanism can also filter out the over-fitting of single meta-classifier and reduce the possibility of over-fitting in the whole model.In addition,this paper presents an improved scheme to control the depth of the tree,conduct pruning operation and reduce the complexity of the model to solve the over-fitting problem in database intrusion detection of single gradient lifting decision tree.In order to validate the performance of the proposed algorithm,nine sets of UCI data were selected for basic classification ability test and KDDCUP99 data sets for intrusion specific test.Gradient Boosting Decision Tree and Random Forest were also used in the test for the control group in the experiment.The experimental results showed that the significance of the Random Forest algorithm based on gradient boosting is improved.Finally,this paper designed the database intrusion detection system with four modules of data acquisition,data preprocessing,intrusion detection unit,alarm response and described every module in detail.