Network intrusion detection with workflow feature definition using bp neural network

Abstract

The major problem of existing intrusion detection using neural network models is recognition of new attacks and low accuracy. The paper describes an intrusion detection method based on workflow feature definition according to KDD cup 99 types with feed forward BP neural network. The workflow can define new attacks sequence to help BP neural network recognize new attacks. The method takes network traffic data to analyze and classify the behaviors of the authorized users and recognize the possible attacks. The experiment results show that the design is effective. © 2009 Springer Berlin Heidelberg.