Information security is one of the important role to protect the information from unauthorized person. Intrusion detection is a classifier to classify the data as normal and various types of attacks. Data mining based decision tree algorithm play very important role to develop the robust IDS to classify the attacks which is harmful for our system. In this research work, used decision tree techniques as classifier to classify the attacks. We have also develop the robust ensemble model which is combination of C4.5, Simple CART and decision tree that gives better accuracy. Our proposed ensemble model gives 99.70% with 80-20% training -testing partition. We have also applied the feature selection technique to computationally increase the performance of model. Our proposed model gives 99.80% in 11 features with info gain feature selection technique while 98.80% in 16 features with gain ratio feature selection technique. Keywords: Intrusion Detection System, Classifier, Attack. I.INTRODUCTION Now a day's data is increasing day by day in every organization. To secure such information is one of the most important issue for every organization. Information security is one of the important issues to protect the information from unauthorized access. An intrusion detection system is used to detect several types of malicious behaviors that can compromise the security and trust of computer system. Classification plays very important role to classify the unwanted data. Actually IDS is a classifier that classifies the different types of attacks and normal data. In this research work, we have used decision tree algorithm to develop the classifier which classify the normal and different types of attacks. Various researcher have worked to develop the IDS. J. Jabez et al. [2] have presented the details of new approach called Outlier Detection approach to detect the intrusion in the computer network. Their training model was consist of big datasets with distributed environment that improves the performance of Intrusion detection system. There proposed approach was also been tested with the KDD data sets that are received from real world. Experimental results shows that proposed IDS system took less execution time and storage to predict and also the performance of proposed IDS is better than that of other existing machine learning approaches and can significantly detect almost all anomaly data in the computer network. Y. Maleh et al. [3] have evaluated the performance of their intrusion detection model using KDDcup'99 database and they studied the variations in detection rate and false positive, when the number of IDS increases in the network. They proposed a hybrid intrusion detection model for WSN. Their IDS used a learning algorithm based on the SVM and a detection technique based on the attack signatures and their combined model of IDS achieved a higher rate of intrusion detection almost 98% with a number very reduces false alarms near 2%. N. Sharma et al. [4] have observed several research works and they have compared t...