基于多层前向神经网络入侵检测系统的研究

针对目前入侵检测系统不能有效检测未知入侵行为的问题，根据神经网络的自学习和自适应性强的特点，采取了将多层前向神经网络与入侵检测系统相结合的方法，提出了一种入侵检测模型，给出了此模型中神经网络模块的改进训练算法，实验证明，此算法入侵检测率可达86％，最大误报率为3％，加大训练样本可进一步提高检测率，从而更有效地检测出未知的入侵行为；此算法实时性强，可有效提高神经网络的学习效率。

Study of Intrusion Detection System Based on Multiplayer Forward Neural Network

Aimed at the problem that intrusion detection system couldn't detect undefined intrusion behavior effectively, according to the self-learning and self-adaptability of the neural network, this paper integrated the neural network with intrusion detection system, proposed a network intrusion detection model and gave an improved algorithm for neural network model. The experiments proved that intrusion detection rate is 86 percent, and maximal error is 3 percent, which can advance detection rate by more training samples, therefore, can detect undefined intrusion behavior effectively. The algorithm is strongly real-time and can improve the study efficiency effectively.