基于粗糙集的进化神经网络入侵检测系统研究

摘要

针对目前实时入侵检测系统所处理的网络数据具有的非线性和高维的特点,提出基于粗糙集理论的进化神经网络入侵检测方法.对网络中截获的数据,利用粗糙集属性约简方法对其属性集进行约简,得到影响分类精度的重要属性.把约简后形成的训练样本进行数值化和归一化处理,作为神经网络的输入数据,再利用遗传算法较强的宏观搜索能力和全局寻优的特点,优化神经网络权值,并在此基础上进行神经网络学习,从而建立入侵检测系统的优化分析模型.实验结果表明,该算法学习速度快,有效提高了入侵检测系统的检测效率.%Research into the network security question of intrusion detection. According to the data of dimensional network intrusion and nonlinear features, in order to improve the network security, is proposed based on rough set and evolutionary neural network intrusion detection methods. At first, using rough set to reduce attribute of network intrusion data, eliminate the dimension of redundant information, simplify the neural network's input, then using genetic algorithm to optimize neural network weights, accelerate neural network learning speed. Finally the neural network model is adopted to optimize the data after rough set and capture network intrusion detection data of nonlinear rule. Experimental results showed that, compared with other network intrusion detection methods, this method is fast learning, detection accuracy and high rate of fail, it is a kind of efficient and real time network intrusion detection methods.