摘要
提出了采用紧致型小波神经网络来构建服务器预警系统,将小波和神经网络直接融合,使网络训练过程从根本上避免了局部最优等非线性优化问题,小波神经元的低相关性,也使得小波神经网络有更快的收敛速度;将服务器中的日志数据数值化后进行网络训练,获得一个基于小波神经网络的入侵分类器。实验结果,表明小波神经网络系统自适应能力强、学习速度快、预警精度高、在入侵检测领域有良好的实用性。
A server warning system based on wavelet neural network is presented. Wavelet neural network is a kind of neural networks, which combines wavelet theory with neural network theory, that is wavelet function forming neuron, it avoid the problem of nonlinear optimizations, such as local optimization. WNN has fast convergence because of the low correlation of wavelet neuron. In this system ,the WNN was trained by using numbered records of system log files in server and the classifier for intrusion detection can be obtained. The experimental result shows that the performance of this system is highly adaptive , the learning speed is fast, and the rate of warning accurate is high, so it is practicable in intrusion detection.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2005年第3期343-346,共4页
Journal of University of Electronic Science and Technology of China
基金
江苏省政府资助项目(BR2003015)
江苏省高校自然科学重点资助项目(03KJA52066)
