Computational-Intelligence Models for Visualization-based Intrusion Detection Systems.

Abstract

Intrusion Detection Systems (IDS’s) are essential components in a network communication infrastructure, as they enforce security by monitoring traffic and detecting malicious activities. In this research, Computational Intelligence models support an IDS technology to obtain a synthetic, effective visualization of the traffic analysis. Auto-Associative Back-Propagation (AABP) neural networks map feature vectors extracted from traffic sources into a compact representation on a 2-D display. During training, the neural network learns to compress the data in an unsupervised fashion; at run time, the trained neural component synthesizes an effective, 2-D representation of the traffic situation. Empirical tests involving Simple Network Management Protocol (SNMP) traffic proved the validity of the approach.