Neural Network Techniques for Improved Intrusion Detection in Communication Systems

In this paper we discuss the main research approaches in the development of automated and systematic methods for intrusion detection. In this critical overview, the key concept underlying the presented intrusion detection systems is that they involve pattern analysis techniques to discover consistent and useful patterns of system features that describe program and user behaviour, and the set of relevant system features to compute and recognize anomalies and known intrusions. The derived patterns compose the inputs of classification systems, which are based on statistical, structural and machine learning pattern recognition techniques. The second goal of this paper is to present a conceptual framework for systematically applying neural network techniques in intrusion detection problems, derived from the constraints put by the critical overview of pattern analysis based intrusion detection systems. The novelty of this work lies on the attempt to define a systematic and overall conceptual framework for applying neural network techniques in the design of such systems by identifying all underlying factors.