摘要
提出一种基于混沌神经元的混合前馈型神经网络,用于检测复杂的网络入侵模式.这种神经网络具有混沌神经元的延时、收集、思维和分类的功能,避免了MLP神经网络仅能识别网络中当前的滥用入侵行为的弱点.对混合网络进行训练后,将该网络用于滥用入侵检测.使用DARPA数据集对该方法进行评估,结果表明该方法可有效地提高对具备延时特性的Probe和DOS入侵的检测性能.
A hybrid feedforward neural network based on chaotic neuron is proposed to detect complicate network intrusion. The proposed neural network has the capability of time-delay, collection, thinking and classification, based on which the weakness of general neural network is avoided which can only detect current misuse intrusion. The neural network is trained and applied to misuse intrusion detection cases. This approach is evaluated by using DARPA data set. Results show that the system's capability of detecting time-delayed Probe and DOS attacks is enhanced effectively by using the proposed approach.
出处
《控制与决策》
EI
CSCD
北大核心
2007年第4期432-435,共4页
Control and Decision
基金
国家自然科学基金项目(60473073)
国家863计划项目(2004AA1Z2060)
国家973计划项目(2006CB303000)
广东省自然科学基金项目(04010589)
关键词
网络安全
入侵检测系统
前馈型神经网络
混沌神经网络
Network security
Intrusion detection system
Feedforward neural network
Chaotic neural network