The growth of internet applications makes the intrusions on the networking system at a high level. In such a situation, it is essential to offer security to the network systems by an efficient intrusion detection and avoidance mechanisms (IDS and IPS). This can be accomplished by formulating an efficient intrusion detecting system that finds effective attributes of the IDS dataset and can distinguish the abnormal and normal activities in the network. Many intrusion detection system applications have been proposed in the past and the systems are having limitations in terms of detection accuracy. To overcome these limitations, the proposed research work offers a new scheme of identifying the suitable attributes for the IDS by prediction mechanisms such as Bayesian classification and support vector machine. The experiments are conducted using KDD CUP’99 dataset and the observation shows that the proposed approach reduces false positive alarms, detects suitable attributes for the intrusion detection, precisely and accurately. Hence, the overall accuracy detection has been improved.

intrusion detection system, prediction, SVM, Bayesian, anomaly detection, misuse detection.