Intrusion Detection System is required to detect the intrusions that are spread through the Network. The main idea of this paper is to reduce the average control path latency incurred between request and response of the system as well as the increasing the detection rate of network attack groups. This paper proposes two approaches to design the efficient and accurate Intrusion Detection System. The proposed system make use of Individual Feature Set and Layer wise approach to achieve the efficiency and accuracy in detecting the network attack groups. The proposed system categorizes the network attacks in to four groups such as Denial of Service attacks, User to Root attacks, Remote to Local attacks and Probe attacks. Experimental results shows that the attack detection rate of the proposed method is high when compared to the other methods such as Support Vector Machine, C4.5 algorithm, Decision Tree with Principle Component Analysis, K means clustering and Multi Classifier for detecting the network attacks.