IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.12, September 2017 1 Manuscript received December 5, 2016 Manuscript revised September 20, 2017 Proposing a Hierarchical Classifier to Detect Attack in Network Intrusion Detection Amin Shahraki Moghaddam 1 , Javad Hosseinkhani 1 , Anoosh Mansouri Birgani 1 , Amirreza Sardarzadeh 2 , Zeynab Sayad Arbabi 1 , and Sadegh Gilani 1 Department of Computer, Zahedan Branch, Islamic Azad University, Zahedan, Iran 1 Zahedan, Iran Department of Computer, Damavand Branch, Islamic Azad University, Damavand, Iran 2 Damavand, Iran Summary The task of intrusion detection system intrusion detection and disclosure practices are responsible. This system monitors network traffic and reports by user activity, detects illegal activities. Detect, identify and classify classes of attacks on computer networks, one of the major challenges in the field of intrusion detection is to determine the type of attack class. Neural networks, support vector machines and Bayesian networks as a classifier to classify and identify the type of attacks are used. Many researches have been conducted using a combination of the classifier. This classifier with putting together several different classifiers to detect attacks that are used to determine the type.be used as it is challenging. The classifier of support vector machine and a neural network classifier to determine the best of each class have detected the attack. And also the best way to arrange those bands that plays a big part in yield is proposed. Simulation results show that the proposed classifier can improve the classification performance better than similar acts. Key words: Intrusion Detection, Support Vector Machine (SVM), Neural Network, Hierarchical Classifier. 1. Introduction The intrusion detection is processing to detect unauthorized attempts to access a network or decrease its performance. In intrusion detection must first understand how the attacks were carried out. Thus by obtained understanding there is a two-step method to stop it. The first one is detecting the pattern of dangerous activities then ensure that those activities are classified in safe categories not attacks category. That's why most of intrusion detection systems rely on a mechanism to update their software to act against network threats fast enough. Of course, intrusion detection alone is not good enough and attack should be followed to track the hacker in order to dealt with him appropriately. Intrusion is the act to violate a security component such as confidentiality, integrity and availability from breach in a system or application. Intrusion detection system has the task of discovering and exposing the attacker's actions. The system detects unauthorized activities by monitoring network traffic and user activity reports. Detection, detecting the type of classification and classification of attacks on computer networks is one of the major challenges in the field of intrusion detection and determining the type of attack class. Neural networks, support vector machines and Bayesian networks as classifier are used to classify and identify the type of attacks. A number of researches were carried out on the field of using these classifiers combined. These classifiers are used by putting together several different classifiers for detecting and determining the type of attack. Classification accuracy, reducing false alarms and increasing appropriate warning rates can be noted as evaluating criteria of an efficient classifier in attack detection and determining its type. Using what type of classifier for detecting and determining the class of a particular attack is always noted as a challenge. In this study the most important challenges and criteria of an efficient classifier in detecting and determining the type of investigated attack and a way to classify and determine the type of attacks is suggested as hierarchically and hybrid. The proposed classifier method for each class, for put attacks consecutively together so that the output of each classifier (Failure to detect attack) would be the entrance of next classifier. By simulations done in MATLAB environment on support vector machines and neural networks classifiers, the very best of each classifier for each class of attack were detected. The best way to arrange classifiers to have an effective performance have suggested. The simulation results indicate that the proposed classifier can act better in improving the classification compared to similar cases. 2. Related Works Prasad et al. [1] presented a model for intrusion detection systems to identify anomalies. This model is based on