With the widespread propagation of Internet of Things through wireless sensor networks, massive amounts of sensor data are being generated at an unprecedented rate, resulting in very large quantities of explicit or implicit information. When analyzing such sensor data, it is of particular importance to detect accurately and efficiently not only individual anomalous behaviors but also anomalous events (i.e. patterns of behaviors). However, most previous work has focused only on detecting anomalies while generally ignoring the correlations between them. Even in approaches that take into account correlations between anomalies, most disregard the fact that the anomaly status of sensor data changes over time. In this article, we propose an unsupervised contextual anomaly detection method in Internet of Things through wireless sensor networks. This method accounts for both a dynamic anomaly status and correlations between anomalies based contextually on their spatial and temporal neighbors. We then demonstrate the effectiveness of the proposed method in an anomaly detection model. The experimental results show that this method can accurately and efficiently detect not only individual anomalies but also anomalous events.

中文翻译：

---

一种基于无线传感器网络的物联网上下文异常检测自适应方法

随着物联网通过无线传感器网络的广泛传播，海量传感器数据正以前所未有的速度产生，从而产生大量显性或隐性信息。在分析此类传感器数据时，不仅要准确有效地检测个体异常行为，还要检测异常事件（即行为模式），这一点尤为重要。然而，以前的大多数工作只关注检测异常，而通常忽略它们之间的相关性。即使在考虑异常之间相关性的方法中，大多数方法也无视传感器数据的异常状态随时间变化的事实。在本文中，我们通过无线传感器网络提出了一种物联网中无监督的上下文异常检测方法。该方法解释了动态异常状态和异常之间基于上下文的空间和时间邻居的相关性。然后，我们证明了所提出方法在异常检测模型中的有效性。实验结果表明，该方法不仅可以准确有效地检测个体异常，还可以检测异常事件。