一种改进的基于神经网络的入侵检测算法

An Improved Intrusion Detection Algorithm Based on Neural Network

径向基神经网络具有良好的分类特性,被广泛应用于入侵检测系统的研究中,然而RBF神经网络需要事先确定隐层神经元数目,并且容易陷入局部最优。利用遗传算法优化隐层神经元数目,并且基于粒子群思想优化隐层到输出层的权值,同时给出了详细的算法流程。经Lincoln实验室入侵检测系统数据评估集合测试,该智能算法的检测成功率大大提高,并且训练时间比较短,完全可以应用于入侵检测系统中。

Radial basis neural network has good classification performance,and is widely usedin the research of network intrusion detection system,however the RBF neural network needs a predetermined number of neurons in hidden layer,and is easy to fall into local optimum.An improved algorithm is worked out,which use genetic algorithm to optimize the number of the hidden neurons.and use Particle Swarm algorithm hidden layer to the output layer weights.A detailed algorithm is gived.This algorithm is tested based on The Lincoln laboratory intrusion detection system data evaluation set.The result shows that this algorithm has greatly improved the success rate,and has short training time.It can be applied to intrusion detection system.