HIDE : a Hierarchical Network Intrusion Detection System Using Statistical Preprocessing and Neural Network Classification

Abstract

Abstract—In this paper we introduce the Hierarchical Intrusion DEtection (HIDE) system, which detects network-based attacks as anomalies using statistical preprocessing and neural network classification. We describe our system architecture and the statistical preprocessing technique and components. We tested five different types of neural network classifiers: Perceptron, Backpropagation (BP), Perceptron-backpropagation-hybrid (PBH), Fuzzy ARTMAP, and Radial-based Function. Our results indicate that BP and PBH provide more efficient classification for our data than the alternatives. We also stress-tested the entire system, which showed that HIDE can reliably detect UDP flooding attacks with attack intensity as low as five to ten percent of background traffic.