Research On Intrusion Detection Technology Based On Variable Precision Rough Set

With the convergence of information and communication technology and the economy, industry and other fields, the corresponding network management issues increasingly prominent, brought new challenges to network security. Intrusion detection as a protect means is especially important. Traditional intrusion detection system exist low-detection rate, real time delay, the high rate of false negatives and false alarm, can't meet the demand of network security. but also requires a lot of prior knowledge to gain experience to structure models can only obtain relatively satisfactory detection performance. For find a method which does not much prior knowledge has practical significance.Intrusion detection process is to distinguish between normal behavior and abnormal behavior. Support vector machine is to some extent as a classifier for the optimal control of a technology that solve limited-small sample, nonlinear and high dimensional pattern recognition exhibit many unique advantage and be able to used in machine learning fields. That's reason the paper introduce the SVM to intrusion detection.Intrusion contains many attributes, if the data is directly as training samples, it will take along time. Of course, not all the attributes which are essential, and some features are imprecise, rough set theory is precisely to deal with these uncertainties, imprecise data has advantages, variable precision rough set overcome the defects of rough set is sensitive to data. Therefore, variable precision rough set on attributes reduction have a better advantage.The paper combine the attributes reduction of variable precision rough set with the classifier of support vector machine applied in intrusion detection system. On the basis of the common intrusion detection framework, proposed intrusion detection model based on variable precision rough sets and support vector machine, and analysis each module of the model. Experimental data sets using KDDCUP99, detection result of the VPRS and SVM compare with other detection methods, show that the rate of detection is nearly to others, but false negatives rate is lower that others. Adopt the measure of variable precision rough sets reduction algorithm to process data, reducing the amount of data storage, training time is shortened, intrusion detection system's real-time is improved.