Abstract: In the growing Era of Information Technology, the use of Internet is the basic need of all, but it also caused new problem called Network Security. To solve these problem Intrusion Detection Systems (IDSs) are used. IDS provide the information related to malicious activities and protect network from vulnerabilities and provide better information security. Intrusion detection system provides security for huge network like Internet. Huge amount of data is transferred on the network these days hence system should analyze and detect the malicious activities. The system should analyze and recognize intrusion in effective way and in less time. The main hurdle in network security is to enhance the intrusion detection system. In this paper we proposed Hybrid anomaly Intrusion Detection System Using Outlier Mining with Support Vector Machine on the basis of TCP transmission control protocol header information and other attributes. Here we are using different approaches from that one is k-mean clustering and another is one-class support vector machine to formulate and model different sessions already presented in the dataset provided by the MIT DARPA 99 dataset. After that we provide the testing set to the model for predicting the attack scenarios of session.

Keywords: IDS,k-means clusterig,one class SVM,outlier minning,TCP.