Research On Feature Selection Of Intrusion Detection Based On Improved GWO Algorithm

Today's Internet is developing rapidly,especially in the era of big data and cloud computing,and people's demand for network security is increasing.And technological innovation is changing with each passing day,and the means of network intrusion are characterized by concealment and silence.Therefore,it is imperative to establish an effective network intrusion prevention mechanism.Nowadays,some mainstream network intrusion detection models are based on the idea of pattern recognition and machine learning.However,with the increase of data magnitude,the efficiency of intrusion detection is getting lower and lower,the effect is getting worse and worse,and the data is massive.These intrusion detection models present challenges.Moreover,as the data dimension increases dramatically,a large amount of irrelevant and redundant information is generated,which can greatly reduce the performance of machine learning algorithms,increase computational complexity,and cause "dimensionality disaster" and "over-fitting" problems.Feature selection is an effective means to solve this problem.In this paper,the network intrusion data set data is selected as the real experimental object,and the feature selection method is used in the intrusion detection.The detection effect is good and the number of feature selection is used as the measurement index.Around these issues,the work done in this article includes:After inquiring about the inefficiency and convergence of traditional intelligent algorithms and the problem of falling into the local optimal solution,after consulting the relevant literature,the recently proposed intelligent algorithm was selected: the grey wolf optimization algorithm as the feature selection algorithm and used in the intrusion detection system.in.At the beginning of each iteration,the algorithm first determines the optimal solution in this iteration,and then updates the solution space by approximating all the alternative solutions in this iteration to achieve the approximate global optimal solution.Aiming at the stagnation problem of traditional binary gray wolf algorithm,this paper proposes a grey wolf optimization algorithm based on distance greedy strategy.This algorithm abandons the traditional threshold binaryization idea and puts the standard of dualization on each one.The superiority and inferiority of the solution itself improves the upper limit of the algorithm effect.After discussing several feature selection methods,several more effective feature selection methods for GWO,EBGWO and DGSBGWO were proposed in the KDDCUP1999 dataset.Experiments show that the gray wolf optimization algorithm using distance greedy strategy is better than traditional intelligent algorithms in different classifiers.