摘要
为了提高入侵检测系统的正确率,降低误检率,对基本的免疫克隆选择算法采用抗体的克隆数目与亲和度成正比且克隆数目线性递减、变异概率线性递减、新的替换策略、变异概率和抗体克隆数量突变进行改进。对抗体克隆策略的改进保证了算法的收敛速度,避免了算法后期的震荡,变异概率的自适应变化加强了算法后期的收敛,新的替换策略、变异概率和抗体克隆数量突变能够有效地避免算法陷入局部最优。经过KDD Cup 1999数据集的训练和检验数据的仿真测试,改进后的算法具有较高的检测正确率和较低的误检率,而且新算法收敛速度快,不易"早熟"。
In order to improve the correct rate of intrusion detection system and reduce false positive rate,on the basic immune clonal selection algorithm using antibody clone number and affinity is proportional to the degree and the number of clones linear decreasing,mutation probability decreasing linearly,newreplacement strategy,the number of mutation probability and antibody clonal mutation for improvement. The improvement of antibody cloning strategy to ensure the convergence speed of the algorithm and avoid the shock of the late. Adaptive changes in the mutation probability strengthen the convergence of the algorithm for the late. Newreplacement strategy and the number of mutation probability and antibody clonal mutations can effectively avoid the algorithm into a local optimum. After the training and testing for Cup KDD 1999 data set,the improved algorithm has the advantages of higher detection accuracy rate and lower false detection rate,with fast convergence speed,and it is not easy to ＂premature＂.
出处
《计算机技术与发展》
2016年第5期86-90,共5页
Computer Technology and Development
基金
陕西省自然科学研究项目(14JK1828)
关键词
入侵检测
克隆选择
变异概率
克隆策略
自适应
intrusion detection
clonal selection
mutation probability
cloning strategy
adaptive
作者简介
牛永洁（1977-），男，硕士，讲师，CCF会员，研究方向为数据挖掘、智能算法、网络安全。
