一种基于改进的BP神经网络的入侵检测方法

Intrusion Detection Method Based on Improved BP Neural Network

针对普通BP神经网络算法学习收敛速度慢、易造成局部极小的问题,提出一种改进的BP神经网络入侵检测方法,其采用拟牛顿的方法进行学习,即对目标矩阵求二阶导数.运用该方法能够有效提高学习速度,消除局部极小.仿真结果表明,改进的BP神经网络入侵检测方法收敛速度快,比标准的BP入侵检测方法误检率低,能够很好地提高学习效率,更加有效地检测攻击行为.

General BP neural network algorithm has a low constringency speed,which will easily result in a local minimization problem.An advanced intrusion detection method based on improved BP neural network is proposed.It adopts the Newtonian method to seek the second rank differential coefficient of the target matrix and can expedite learning speed rapidly and clear up part particle.Simulation results demonstrate that the intrusion detection method based on improved BP neural network,with a higher constringency speed and a lower false-detection rate,can improve learning efficiency and inspect attacking behaviors effectively.