AFSA-BPNN在网络安全中的应用研究

Application of Back Propagation Neural Network Optimizing by Artificial Fish Swarm Algorithm in Network Intrusion Detection

为了有效地检测网络入侵行为,提出一种人工鱼群(AFSA)算法优化BP神经网络(BPNN)入侵检测模型。首先将BP神经网络的权值和阈值编码为AFSA的人工鱼状态,然后通过人工鱼群觅食、聚群、追尾等行为,对BP神经网络的参数进行优化,寻找到BP神经网络的最优参数,最后利用建立的最优BP神经网络模型,对网络入侵行为进行检测。在Windows XP操作系统,Matlab 2012平台上,采用KDD CUP 99数据集仿真测试,相对于传统的BP神经网络模型,本文模型可以显著提高网络入侵检测正确率,有着更加广泛的应用前景。

In order to detect the network intrusion effectively,an artificial fish swarm（ AFSA） algorithm is proposed to optimize the BP neural network（ BPNN） intrusion detection model. In this paper,firstly,weights and threshold coding state of BP neural network is artificial fish of the AFSA,followed by artificial fish swarm foraging,Poly Group,rear end,which is used to optimize the parameters of BP neural network,to find the optimal parameters of the BP neural network; After that,using the establishment of optimal BP neural network model,the behavior of the network intrusion are detected. In the Windows XP operating system,on the platform of MATLAB 2012 by the KDD cup 99 data set of simulation test,compared with the traditional BP neural network model,the proposed model can significantly improve the network intrusion detection accuracy and has a more extensive application prospects.