This paper proposes a hybrid technique for intrusion detection model using K-means clustering, attribute selection and decision tree. K-means clustering is a very simple and convenient clustering method when it comes to grouping anomalies and the different attack types in network traffic. An enhanced mechanism is developed using the Cluster center initialization algorithm for k-means clustering and decision trees using the entropy method. After the clustering is done, attribute subset selection is done using entropy method and final classification of attack categories is done using decision trees. It works in two modes: online and offline. Offline mode works on the sample data which is processed to obtain the rule set of the decision tree. The data from the online mode is then compared against those rules to determine their category and identify the intrusions in the packet. © 2006-2015 Asian Research Publishing Network (ARPN).