基于隐私保护的分布式异常检测方法

为获得鲁棒性的全局异常检测模型,需要多个组织之间的知识共享.存在的分布式异常检测技术常基于原始数据的交换或共享,侵犯了各自的隐私权,令人难以接受.基于隐私保护的分布式异常检测方法,采用本地模型共享技术,在保证数据隐私性的同时完成全局异常检测任务.通过7种异常检测模型在仿真和真实数据集上的实验说明,所提出的方法在保护数据隐私的同时,其全局异常检测效果能接近甚至超过将所有数据集中后建立的全局模型.

Privacy preserving distributed anomaly detection method

To achieve robust global anomaly detection models, different companies or organizations should share their
knowledge of data. However, the sharing of production data will lead to violation of privacy. It is unaccepted to co-operate
with the risk of disclose private or sensitive data. The existing distributed anomaly detection techniques always neglect
the requirement and are based on the sharing or exchanging of production data. The proposed privacy preserving distributed
anomaly detection method employs local model sharing technology to preserve the privacy of data. Mean while, the proposed
method has comparable or even better performance on the synthetic as well as several real life data sets by seven different
anomaly detection models.