Research On Bayesian Intrusion Detection Based On Improved Genetic Algorithm

As a supplement of firewall,Intrusion Detection System can proactively defend the way to predict the network environment in advance of the security risks and timely response,Assist the firewall to ensure network security.The core technology of intrusion detection system lies in the detection algorithm,efficient and stable detection algorithm can accurately identify the connection with less characteristic data and advanced warning.This thesis focuses on the detection algorithm and study the intrusion detection system on the basis of predecessors' research.The main work is as follows:(1)Introduce the technical background of intrusion detection system,the research status and the purpose and method of this thesis.Then,introduce the intrusion detection KDD99 data set especially the method of data preprocessing in detail and use the Bayesian algorithm to classify the data set and analyze the results in detail.(2)Aiming at the Bayesian classification result and the high dimensionality characteristic of data set,puts forward the genetic algorithm for feature selection in order to obtain better classification performance.When choosing the best individual,consider the variables of fitness function,we propose an individual evaluation method for the lake of elite preservation strategy.This method does not use the fitness function as the only criterion but rather considers the factors that affect the individual's optimal and the influence factors are given different weightings according to the importance degree,and put forward the new standard of individual evaluation based on the fitness function.Introduced the concept of gene similarity in the process of cross manipulation,compared the similarity of individual genes,and the close relatives of gene with the similarity of the threshold were prevented from crossing to improve the individual diversity of the population.In the process of mutation,in order to accelerate the convergence rate,using the dynamic nonlinear mutation operator.This mode of operation can speed up the convergence rate while ensuring the randomness of evolution,which is more consistent with the theory of genetics.(3)The improved genetic algorithm proposed in this thesis is used to select the KDD99 data set,and the selected feature subset is classified by Bayesian algorithm.To verify the effectiveness of the proposed method,we designed three different contrast experiments.The results show that the improved genetic algorithm can obtain a lower data dimension and improve the detection rate.