Hybrid Model Based on Artificial Immune System and PCA Neural Networks for Intrusion Detection

Abstract

Intrusion detection systems (IDS) are developing very rapid in recent years. But most traditional IDS can only detect either misuse or anomaly attacks. In this paper, we propose a method combining artificial immune technique and Principal Components Analysis (PCA) neural networks to construct an intrusion detection model capable of both anomaly detection and misuse detection. Initially an artificial immune system detects anomalous network connections. In order to attain more detailed information about an intrusion, PCA is applied for classification and neural networks are used for online computing. The experiments and evaluations of the proposed method were performed with the KDD Cup 99 intrusion detection dataset, which have information on computer network, during normal behavior and intrusive behavior. Results indicate the high detection accuracy for intrusion attacks and low false alarm rate of the reliable system.