基于神经网络的程序异常监测

Program behavior anomaly detection based on neural network

针对传统入侵检测系统的不足,研究了基于反向传播神经网络的程序异常检测方法,提出了一个改进的利用多层前馈网络的预测功能和异常区域判定方法检测系统异常的算法.详细讨论了算法的基本原理、数学基础、设计和实现方法.通过实验,分析算法的优缺点,验证了算法的可行性和有效性.

Neural network can be used in anomaly detection.To improve the traditional intrusion detection system performance,we often have to change the neural network structure and detection algorism.And because intrusion techniques are most changeable and unpredictable,it is therefore impossible to always use the fixed detection techniques to catch exactly all possible intrusions.In this paper,an improved algorithm of process anomaly detection based on back propagation(BP) neural network is proposed,which uses the forecast function of multi-level perception and anomaly area estimating method to detect system anomaly.Some details and issues on the design and implementation of the algorithm are discussed.The experimental results are also illustrated and analyzed.