Outlier detection in the Internet of Things (IoT) is an essential challenge issue studied in numerous fields, including fraud monitoring, intrusion detection, secure localization, trust management, and so on. Conventional outlier detection technologies cannot be used directly in IoT due to the open nature of wireless communication as well as the resource-constrained characteristics of end nodes. Therefore, this article provides a comprehensive survey of new outlier detection approaches based on machine learning for IoT. The approaches are first carefully discussed based on their adopted machine learning algorithms. In addition, the performance of them with respect to the advantages and the drawbacks are compared in detail, which naturally leads to some open research issues that are analyzed afterward.

中文翻译：

---

物联网中基于机器学习的异常值检测方法

物联网（IoT）中的异常检测是在众多领域研究的必不可少的挑战问题，包括欺诈监控，入侵检测，安全本地化，信任管理等。由于无线通信的开放性以及端节点的资源受限特性，传统的离群值检测技术无法直接在物联网中使用。因此，本文对基于物联网的机器学习的新异常检测方法进行了全面调查。首先根据所采用的机器学习算法对这些方法进行仔细讨论。另外，对它们在优缺点方面的性能进行了详细比较，这自然导致了一些未解决的研究问题，随后将对其进行分析。