基于神经网络的入侵检测系统的设计与实现

摘要

Since traditional IDS (intrusion detection system) can not identify the unknown attacks,to make up the defect of it,an intrusion detection system based on BP neural network and honeypot (BPIDS) is designed and implemented in this paper.The system contains a two-phase detection model,the perceptron detection model applying the perceptron learning method and the BP network detection model applying BP neural network respectively.The perceptron detection model is used to distinguish the attack class from the normal classes,while the other model focuses on identifying some specific attack types on that basis.At last,an experiment is designed to test the detection capability of BPIDS.The results of the experiment show that the BPIDS has a better detection rate and a lower false alarm rate on intrusion activities in monitoring network.%由于传统的入侵检测系统无法识别未知攻击,为了弥补其不足,设计和实现一个基于蜜罐和BP神经网络的入侵检测系统BPIDS.该系统包含两阶段检测模型,它们分别是应用感知器学习方法的感知器检测模型和应用BP神经网络的BP网络检测模型.其中感知器检测模型用于划分正常类和攻击类,而BP网络检测模型则在此基础上对一些具体的攻击类型进行识别.最后,设计实验对BPIDS的检测能力进行测试.实验结果表明,BPIDS对被监控网络中的入侵行为具有较好的检测率和较低的误报率.