Identification of effective network features to detect Smurf attacks

Abstract

Intrusion detection system (IDS) detects intrusion attempts on computer systems. In intrusion detection systems, feature reduction, feature extraction and feature selection play important role in a sense of improving classification accuracy while keeping the computational complexity at minimum. Smurf attack is one of the common denial-of-service attack methods. In this paper, principal component analysis method is used for feature selection and dimension reduction. TCP dump from DARPA98 dataset is used for the experiments. 32 basic features are extracted for the selection of effective features in TCP/IP header to detect Smurf attacks.