The Application Of Support Vector Machine Based On Grid Search In Network Intrusion Detection

Computer network and computer technology has greatly promoted the development and progress of human society. Computer network is often attacked by hackers and malicious programs due to some defects that exist in computer technology and network protocol. Such attack not only causes the loss of people’s property, but also leads to the leakage of state’s secrets, so the network security problem is becoming the focus of attention increasingly. Compared with anti-virus software and firewall technology, network intrusion detection is a kind of active defense. It can wipe out the action that is destructive before the appearance of damage, so the technology of network intrusion detection has become a new research hot spot.The theory of Support Vector Machine(SVM) is developed on the basis of Statistical Learning Theory,Machine Learning and Statistical Theory form the supervision of this learning model together. Now the algorithm of Support Vector Machine theory mainly aims at the optimization of kernel function parameters and selection of kernel function algorithm. SVM have unique advantage in tacking small sample and high dimension data sets. The data of network communication belongs to the high dimensional data, so using SVM theory for network intrusion-detection is a hot-spot in the present study.To get the optimal parameter combination, the grid search algorithm for optimization of SVM’s parameter gets through traversing kernel functional parameters and punish coefficient. Improved grid search algorithm increases the efficiency of intrusive detection through length of optimized choice that based on grid search. This paper designs a model which uses an improved grid search algorithm to optimize the kernel function of SVM, and the model was applied to incursive detection of network. models of intrusive detection can be divided into data acquisition, data preprocessing module, contrastive module that uses optimized algorithm, model training module and data detective module.This dissertation summarizes the research status of intrusion detection, and analyzes the optimal principle and process for SVM parameters which uses the grid search algorithm and the improved grid search algorithm. This dissertation does a large number of experiments according to different algorithms in intrusive detection, and it proves that optimization for SVM’s parameters that based on improved grid search algorithm has the highest classification accuracy and the least time.