tailieunhanh - Towards a new architecture of detecting networks intrusion based on neural network

The experimentation led us to observe that, depending on the type of the attack, some of the attributes have a marginal effect, although in the opposite, some other have a dominant effect. We have done a comparative study of the solution proposed and the others in the literature. It appears that our solution is more efficient on certain type of attacks. The NSL-KDD dataset has been used to train, test and evaluate our architecture. | International Journal of Computer Networks and Communications Security VOL. 5, NO. 1, JANUARY 2017, 7–14 Available online at: E-ISSN 2308-9830 (Online)/ ISSN 2410-0595 (Print) Towards A New Architecture of Detecting Networks Intrusion Based on Neural Network Berlin H. LEKAGNING DJIONANG1 and Dr. Gilbert TINDO2 1, 2 University of Yaoundé I, Faculty of science, Yaoundé, Cameroon 1 dberlinherve@ ABSTRACT Networks intrusion detection systems appear nowadays as one of the most efficient solution for detecting illegal or suspicious activities in a network. Using neural networks to meet this objective has been studied by many authors. Most of the solutions provided in the literature face the problem of relevance and reliability. One of the major reasons of such a situation is the misconception of a correct profile. In this paper, a modular architecture is proposed, in which each module is dedicated to detecting a particular type of attack. Firstly, each module is trained with all attributes, then secondly some of the attributes used for training are pruned from the training set. This modularity allows us to know which of the system’s module threaten the NIDS performance. The experimentation led us to observe that, depending on the type of the attack, some of the attributes have a marginal effect, although in the opposite, some other have a dominant effect. We have done a comparative study of the solution proposed and the others in the literature. It appears that our solution is more efficient on certain type of attacks. The NSL-KDD dataset has been used to train, test and evaluate our architecture. Keywords:NIDS, Neural networks, MLP, NSL-KDD Dataset. 1 INTRODUCTION The advent of networks has allowed a wide range of services. These services are subject to many attacks and thus security mechanisms have become a necessity. Network intrusion detection systems (NIDS) are one of the most used mechanisms nowadays to detect .