A cascading layered approach for Intrusion Detection to improve detection accuracy of Major and Minor Attacks

Abstract

The growth of computer network for inter communication and data sharing has also raised the challenge of security of data over network. There are many threat or attacks in network that tries to privacy theft of data over network. Intrusion Detection System (IDS) is helpful to protect network form attacks. The attacks over network is divided into four categories i.e. User-to-Root (U2R), Root-to-User (R2L), Prob and Denial-of-Service (DoS). In this paper, a cascading layered based approach is follow to make an IDS. The approach is followed to improve the detection rate of all categories of attacks. Feature selection for Prob & DoS attacks and U2R & R2L attacks is done. An Artificial Neural Network (ANN) is used to improve the detection accuracy of U2R and R2L attacks. The training data for minor attacks is resampled to detect any novel attacks. The proposed approach gives a very good results.