查询结果:   陈超,曹晓梅.改进差分进化算法优化BP神经网络用于入侵检测[J].计算机应用与软件,2018,35(4):310 - 316,324.
中文标题
改进差分进化算法优化BP神经网络用于入侵检测
发表栏目
安全技术
摘要点击数
1085
英文标题
IMPROVED DIFFERENTIAL EVOLUTION ALGORITHM TO OPTIMIZE BP NEURAL NETWORK FOR INTRUSION DETECTION
作 者
陈超 曹晓梅 Chen Chao Cao Xiaomei
作者单位
南京邮电大学计算机与软件学院 江苏 南京 210023     
英文单位
School of Computer and Software,Nanjing University of Posts and Telecommunications,Nanjing 210023,Jiangsu,China     
关键词
BP神经网络 差分进化算法 模拟退火算法 全局寻优 入侵检测
Keywords
BP neural network Differential evolution algorithm Simulated annealing algorithm Global search Intrusion detection
基金项目
国家自然科学基金项目(61202353);国家重点基础研究发展计划项目(2011CB302903);江苏高校优势学科建设工程项目(yx002001)
作者资料
陈超,硕士生,主研领域:网络与信息安全。曹晓梅,副教授。 。
文章摘要
为解决BP神经网络应用于入侵检测时检测率较低、训练时间过长的问题,对改进差分进化算法(SAMDE)优化BP神经网络并用于入侵检测的可行性进行研究。该算法引入模拟退火算法(SA)和一种融合DE/rand/1与DE/best/1的变异算子对差分进化算法进行改进以提高其全局寻优能力。用改进后的算法优化BP神经网络权值阈值。通过逐次的迭代训练使BP神经网络收敛,将优化过的BP神经网络用于入侵检测。仿真实验结果显示,优化的BP网络在收敛速度和精度方面有明显提升,用于入侵检测时提高了检测准确率,缩短了训练时间。
Abstract
In order to solve the problem of low detection rate and long training time when BP neural network is used in intrusion detection, the feasibility of improving BP neural network by using improved differential evolution algorithm (SAMDE) and intrusion detection was studied. This algorithm introduced a simulated annealing algorithm (SA) and a mutation operator that combined DE/rand/1 and DE/best/1 to modify the differential evolution algorithm to improve its global search ability. The improved algorithm was used to optimize the weight threshold of BP neural network. The BP neural network was converged by successive iterative training, and the optimized BP neural network was used in intrusion detection. Simulation results show that the optimized BP network has obvious improvement in convergence speed and accuracy. When used in intrusion detection, it improves the detection accuracy and shortened the training time.
下载PDF全文