Real-Time Network Intrusion Detection System Based on Neural Networks

Traditional Network Intrusion Detection Systems (NIDSs) use rules to detect intrusions, with these rules being updated manually by knowledgeable engineers. With today’s complex network environment, a new systematic method is desired to detect intrusions automatically. Data mining techniques can be used to add a systematic intrusion detection capability to NIDSs. Data Mining is concerned with uncovering patterns, associations, changes, anomalies, and statistically significant structures and events in data. This paper describes a Neural Network (NN) based NIDS architecture. We introduce our experiment on the data available from the 1998 DARPA Intrusion Detection Evaluation. We also present a real-time implementation of a NN based NIDS for Denial of Service (DoS) attacks using open source software.