Deep Learning Approaches to Intrusion Detection
Nassima Chaibi, Baghdad Atmani, Mostéfa Mokaddem
Abstract
In recent years, Cyber Security has become a major concern in scientific research. Machine Learning (ML) and Deep Learning (DL) methods detect intrusions and attacks on the network by predicting risk using data training. DL methods have proved to be more accurate than other network intrusion detection systems. In this paper, we propose an architecture implemented as two methodologies: in the first one, we apply Artificial Neural Network (ANN) and Recurrent Neural Network (RNN) with information gain (IG) as a feature selection method and in the second one using an RNN applying information gain (IG), grain ratio (GR) and correlation attribute (CA) as feature selection method. These two methodologies use the NSL-KDD data-set. Their performance is proven by comparing our results with other previous results. Our results show that RNN outperforms ANN in the two methodologies, while ANN outperforms other machine learning classifiers.