基于机器学习的僵尸物联网检测方法性能评价研究

Performance Evaluation of ZombieIoT Detection Method Based on Machine Learning

机器学习方法已被逐渐应用于僵尸物联网的检测过程。然而,由于已有研究工作所用的机器学习方法以及实验用的数据集不同,很难对不同机器学习方法在僵尸物联网检测的性能进行一致性地评价。为解决上述问题,从是否使用特征降维、使用特征降维时所用的特征维度数量、测试和训练样本分布比例等三个方面设计了不同的实验方案,并在公共数据集上综合评价了多种基于机器学习的僵尸物联网检测方法。实验结果表明,在同样的数据集下,基于LDA-RandomFores的僵尸网络检测算法最好,基于GaussianNB的检测算法性能最差。实验结果为僵尸物联网检测方法研究与应用,选择合适的检测算法,提供了数据支持。

Machine learning methods have been gradually applied to the detection process of zombie IoT.However,due to different machine learning methods and experimental data sets used in existing research work,it is difficult to consistently evaluate the performance of different machine learning methods in the detection of zombie IoT.In order to solve the above problems,different scenarios are designed from three aspects:the feature dimensionality reduction method,the number of feature dimension when using the feature dimensionality reduction method,and the distribution ratio of the test and training samples.This paper comprehensively evaluates multiple scenarios on the public data set.The experimental results show that under the same data set,the performance of botnet detection algorithm based on LDA-Random Forest is the best,while the detection algorithm based on GaussianNB has the worst performance.The experimental results provide data support for the research and application of zombie IoT detection methods,selecting appropriate detection algorithms.