Recurrent nonsymmetric deep auto encoder approach for network intrusion detection system

Abstract

An important part of network security is a network intrusion detection system (NIDS). In the face of the need for new networks, there are issues regarding the feasibility of traditional approaches. More directly, these difficulties are connected to the increasing degrees of human contact required and the diminishing levels of detection precision. A new deep learning intrusion detection approach is presented in this research to overcome these problems. The recurrent non-symmetric deep autoencoder we've suggested for learning unsupervised features is described here (RNDAE). A new deep learning classification model based on LightGBM RNDAEs is also shown. NSL-KDD, CICIDS2017, and CSECICIDS2018 datasets were used to evaluate our proposed classifier in TensorFlow. If our model holds up, it has the potential to be used in the latest generation of network intrusion detection systems (NIDS).

Keywords

• Deep learning
• Anomaly detection
• Autoencoders
• NSLKDD
• CICIDS
• Network security