摘要
为解决神经网络检测方法中检测器需要定期更新、未知攻击检测性能低等问题,利用人工独特型网络的记忆、学习和动态调整能力实现入侵检测.提出一种可用作检测器的多变异模式人工独特型网络,并根据免疫响应原理设计检测算法,使检测器能实时学习新行为特征.仿真结果表明,多变异模式独特型网络检测方法与多层感知器检测方法相比,平均误报率下降了17.43%,未知攻击的平均检测准确率提高了24.17%.
To overcome defects existing in methods based on neural networks, such as the periodical update on detectors and poor performance on unknown attacks, the memory, learning and dynamic regulating abilities of artificial idiotypic networks are used to implement intrusion detection approaches. A multi-mutation-pattern artificial idiotypic network is presented to be used as detectors. By utilizing the immune response principle, the detection algorithm is designed. New behavior features are learnt by detectors in real-time. The detection approach based on multi-mutation-pattern artificial idiotypic network is compared with the detection approach based on multilayer perceptrons through simulations. The results show that the average false positive rate is decreased by 17.43 % and the average detection accuracy of unknown attacks is increased by 24.27 %.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2006年第9期809-812,共4页
Transactions of Beijing Institute of Technology
基金
国家部委预研基金资助项目(YJ0467011)
北京理工大学基础研究基金资助项目(BITUBF200501F4206)
关键词
入侵检测
免疫网络理论
人工独特型网络
intrusion detection
immune networks theory
artificial idiotypic networks
