Anomaly Intrusion Detection Based on Recurrent Neural Networks

Abstract

Security is the main issue within computer networks. Intrusion Detection Systems (IDS) are major ways to guarantee information security and to identify attacks before causing any harm. As a reasonable supplement of the firewall, intrusion detection technology can assist the system to deal with attacks and intrusions. There are numerous problems with the existing intrusion detection systems (IDSs) like the inability to detect unknown attacks and too many false positive rates.  So, this work was suggested to implement IDS based on Recursive Feature Elimination (RFE) methodology to select features and to use Recurrent Neural Networks (RNN) for classification. RNN was used in the classifications for ten classes such as Fuzzers, DoS, Backdoors, Exploits, Analysis, Generic, Reconnaissance, Shellcode, Worms and Normal. The proposed work has achieved a high accuracy of 98%.