基于SOA-SVM的网络入侵检测方法研究

Research on SOA-SVM method based on network intrusion detection

针对网络入侵检测模型的正确率和有效性问题,将人群搜索算法收敛精度高的特点与支持向量机结构风险小、小样本下分类准确率高的优点相结合,提出一种基于人群搜索算法(SOA)和支持向量机(SVM)的网络入侵检测方法(SOA-SVM),该方法将SVM惩罚因子和核函数参数作为人群搜索算法适应度,采用随机搜索和模糊推理方式进行全局寻优,从而找到SVM最优参数并构建入侵检测模型。采用KDD CUP 99数据集进行性能测试,结果表明,SOA-SVM入侵检测模型准确率高,漏报率和虚警率低,在小训练样本情况下依旧具有优良的效果,从而验证了该方法的有效性与稳定性。

According to the problem of accuracy and effectiveness for network intrusion detection model, putting the high convergence precision of searching algorithm into support vector machine(SVM) structure under the risk of small, small sample classification combined the advantage of high accuracy, a population-based search algorithm(SOA) and support vector machine(SVM)method of network intrusion detection(SOA- SVM), this method will punish factor SVM and kernel function parameters as fitness crowd search algorithm, by using random search and fuzzy reasoning method for global optimization, so as to find the optimum parameters and SVM intrusion detection model was constructed. Using KDD CUP 99 data sets, performance testing, the results show that the SOA- SVM intrusion detection model accuracy is high, non-response rates and low false alarm rate, in the case of small training sample still has good effect, so as to verify the effectiveness of the proposed method and stability.