An Efficient NIDS Using Ensemble Techniques for Multinominal Classification

Intrusion detection is one of the major concerns in network administration and security. An intrusion refers to an attack on a system through the network. There is a need to safeguard networks from known vulnerabilities and detect new and unseen, but possible, system abuses. This is done by developing more reliable and efficient intrusion detection systems. The system must detect attacks accurately with the minimum number of false alarms (wrong detections) to be reliable and more efficient. Therefore, a NIDS is developed to improve the accuracy of intrusion detection. Despite high network traffic, this network intrusion detection system efficiently identifies threats. By using various Ensemble techniques and Data mining techniques of classification, the NIDS is developed. In addition to identifying the intrusion, the system also specifies the type of attack. DOS, PROBE, U2R, and R2L attacks can be detected by the developed system. The dataset used is the benchmark NSL-KDD dataset. In the dataset examined, the full NSL-KDD set including attack-type labels and difficulty level is described. Using various attributes of the dataset, by training the model, intrusion detection can be done and the attack can be identified.