“Error back propagation based Recurrent Neural Networks for Intrusion Detection system”

Abstract

Internet usage is increasing daily, which adds to the system's vulnerability. Network security is always a major problem for network administrators and is constantly changing due to the additional application space and requirements of a smart and efficient network. Simple and more efficient software tools include victims on the security side of the protocol that hackers use to perform various types of attacks on the network. The purpose of this review is to establish and coordinate processes to prevent one member against new and known attacks, and to act as a separate security system or an independent network. The neural network connects the body's immune system to receive memory, errors, and synchronized learning. This paper discusses interrelated processes that are largely based on the application of artificial intelligence for intrusion detection and learning processes.Proposed approach finds the type of session i.e. either normal or intrusion where if intrusion found than class of intrusion was detected. Here artificial neural network was used for finding the patterns in the input data. In this work Back propagation is used for the ANN in recursive manner. Proposed algorithm gives an effective framework which has been a promising one for distinguishing interruption of various kind where, one can get the detail of the class of attack also. Experiment has been conduced on real data set where various set of testing data were pass for comparison on different evaluation parameters. Proposed approach detects all sort of attacks applied on the network such as DoS (Denial of service), (R2L) Remote to local, (U2R) User to remote, Probe etc. In this work a Random Forest Tree is used for the detection of intrusion in network. Proposed approach improved 6.87% accuracy, 12.06% Precision and 1.15% recall.