Integrated with cutting-edge equipment and technologies, smart grid takes prominent advantages over traditional power systems. However, hardware and software techniques also bring smart grid numerous security concerns, especially various cyberattacks. Malicious users can launch cyberattacks to tamper with smart meters anytime and anywhere, mainly for the purpose of stealing electricity. This makes electricity theft much easier to commit and more difficult to detect. Researchers have devised many approaches to identify malicious users. However, these approaches suffer from either poor accuracy or expensive cost of deploying monitoring devices. This paper aims to locate malicious users using a limited number of monitoring devices (called inspectors) within the shortest detection time. Before inspectors conduct any inspection, suspicions that users steal electricity are comprehensively assessed, mainly through analyzing prior records of electricity theft as well as deviations between the reported and predicted normal consumptions. On the basis of these suspicions, we further propose a suspicion assessment-based inspection (SAI) algorithm, in which the users with the highest suspicions will be first probed individually. Then, the other users will be probed by a binary tree-based inspection strategy. The binary tree is built according to users' suspicions. The inspection order of the nodes on the binary tree is also determined by the suspicions. The experiment results show that the SAI algorithm outperforms the existing methods.

中文翻译：

---

SAI：一种基于怀疑评估的检查算法，可检测智能电网中的恶意用户

与先进的设备和技术相集成，智能电网具有优于传统电力系统的显着优势。但是，硬件和软件技术也给智能电网带来了许多安全问题，尤其是各种网络攻击。恶意用户可以随时随地发起网络攻击，以篡改智能电表，主要是为了窃电。这使偷电行为更容易实施，也更难以发现。研究人员设计了许多方法来识别恶意用户。但是，这些方法要么准确性差，要么部署监视设备的成本高昂。本文旨在在最短的检测时间内使用数量有限的监视设备（称为检查器）来定位恶意用户。在检查员进行检查之前，全面评估了对用户窃电的怀疑，主要是通过分析先前的盗窃记录以及报告的和预计的正常消耗量之间的偏差来进行的。在这些怀疑的基础上，我们进一步提出了一种基于怀疑评估的检查（SAI）算法，其中将首先对具有最高怀疑的用户进行单独调查。然后，将通过基于二叉树的检查策略来探查其他用户。根据用户的怀疑来构建二叉树。二叉树上节点的检查顺序也由怀疑决定。实验结果表明，该算法优于现有方法。主要是通过分析以前的盗电记录以及报告的和预计的正常消耗量之间的偏差。在这些怀疑的基础上，我们进一步提出了一种基于怀疑评估的检查（SAI）算法，其中将首先对具有最高怀疑的用户进行单独调查。然后，将通过基于二叉树的检查策略来探查其他用户。根据用户的怀疑来构建二叉树。二叉树上节点的检查顺序也由怀疑决定。实验结果表明，该算法优于现有方法。主要是通过分析以前的盗电记录以及报告的和预计的正常消费之间的偏差来进行的。在这些怀疑的基础上，我们进一步提出了一种基于怀疑评估的检查（SAI）算法，其中将首先对具有最高怀疑的用户进行单独调查。然后，将通过基于二叉树的检查策略来探查其他用户。根据用户的怀疑来构建二叉树。二叉树上节点的检查顺序也由怀疑决定。实验结果表明，该算法优于现有方法。其中首先会对怀疑程度最高的用户进行单独调查。然后，将通过基于二叉树的检查策略来探查其他用户。根据用户的怀疑来构建二叉树。二叉树上节点的检查顺序也由怀疑决定。实验结果表明，该算法优于现有方法。其中首先会对怀疑程度最高的用户进行单独调查。然后，将通过基于二叉树的检查策略来探查其他用户。根据用户的怀疑来构建二叉树。二叉树上节点的检查顺序也由怀疑决定。实验结果表明，该算法优于现有方法。