Svm-based Ant Colony Clustering Algorithm In Intrusion Detection Applications

In the high-speed development information age, with the continuous development of network technology and internet sacle,the information security has become one of the most important issues of global.Intrusion detection techonology is the next generation of security technology.It monitors the computers or network,and analysis,check out the intrusions thereby enhancing the security of the system.In the real network environment, Intrusion detection system is difficult to obtain a more complete training data.Active learning support vector machine(SVM) algorithm can solve the training problem of small sample,Greatly reduce the required number of traning data,Intrusion detection system can improve the performance of the classifier.Meanwhile,intrusion detection need to change the external environment constanly updated.As retraining the model takes a long time,retraining all the data for the new model is unrealistic.Therefore,there is a mechanism to generate an adaptive model with the the old model and new information to be updated. Self-organizing adaptive ant colony algorithm has been used in the model, It does not require re-training all data for updateing model.This paper presents a SVM algorithm based on ant colony clustering, the query strategy of SVM active learning algorithm is changed to be basing clustering method.The self-organized ant colony clustering process is divided into several sub-phase,while modify the action mechanism of the ants,to cluster with specific object.The self-organizing ant colony clustering algorithm is used as the query strategy in SVM. Combine well with the two algorithm,and the improved algorithm is applied to intrusion detection system.Evaluate the performance of intrusion detection system with KDDCUP99 data set.Experiment results show that the SVM based on the clustering of ant colony comparision with the pure SVM,reducing the running time,with the self-orgnization, a higher detection rate and low false alarm rate.Meanwhile,comparing with the pure ant colony clustering algorithm,a more efficiency.