Network security field had gained research community attention in the last decade due to its growing importance. This paper addresses directly one vital problem in that field is “Intrusion Detection System” (IDS). As much as many researchers tackle this problem, many challenges arise while converting this research to reliable automatic system. The biggest challenge is to make the system works with low false alarm with new unseen threats. In this paper, we address this challenge by building a descriptive model using different models of deep Recurrent Neural Network (RNNs). (RNN) models has the ability to generalize the knowledge that can be used to identify seen and unseen threats. This generalization comes from RNN capabilities to define in its terms the normal behavior and the deviation accepted to be normal. Four different models of RNN were tested on a benchmark dataset, NSL-KDD, which is a standard test dataset for network intrusion. The proposed system showed superiority over other previously developed systems according to the standard measurements: accuracy, recall, precision and f-measure.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
You must read and accept the copyright terms and conditions(click here for copyright terms).