摘要
基于已知数据的机器学习模型在实际异常流量检测任务中不完全可靠,为此,将不同分布的流量分别作为源域和目标域,建立跨域网络异常流量检测框架,提出了基于联合分布适配的迁移学习方法.通过寻找最优变换矩阵、适配源域与目标域之间的条件概率和边缘概率,实现源域与目标域间的特征迁移,从而解决由于源域与目标域分布差异大所引起的检测准确率下降等问题.实验结果表明,所提方法可以显著提升跨域流量的检测准确率.
In order to solve the problem that the machine learning model based on known data is not completely reliable in actual abnormal traffic detection tasks due to the dynamics of the network environment.The different distributed traffic as the source domain and target domain is used to establish a cross-domain framework for abnormal network traffic detection. The transfer learning method based on joint distribution adaptation is proposed by finding the optimal transformation matrix,adapting the conditional probability and edge probability between the source domain and the target domain,the feature transfer between the source domain and the target domain is realized thereby for solving the problem of the large difference in the distribution of the source domain and the target domain causes problems such as decreased detection accuracy. Experiments show that the proposed method can significantly improve the detection accuracy of cross-domain traffic.
作者
彭雨荷
陈翔
陈双武
杨坚
PENG Yu-he;CHEN Xiang;CHEN Shuang-wu;YANG Jian(School of Information Science and Technology,University of Science and Technology of China,Hefei 230026,China)
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
2021年第2期33-39,共7页
Journal of Beijing University of Posts and Telecommunications
基金
国家重点研发计划项目(2018YFF01012200)
中央高校基本科研业务费专项资金项目(WK2100000009)
安徽省自然科学基金项目(1908085QF266)。
关键词
异常流量检测
跨域
迁移
联合分布适配
机器学习
abnormal traffic detection
cross-domain
transformation
joint distribution adaptation
machine learning