Multihypothesis Sequential Testing for Illegitimate Access and Collision-Based Attack Detection in Wireless IoT Networks
Multihypothesis Sequential Testing for Illegitimate Access and Collision-Based Attack Detection in Wireless IoT Networks
复制标题
无线物联网网络中非法访问和基于冲突的攻击检测的多假设顺序测试
DOI：
--
复制
发表时间：
2021
影响因子：
10.6
通讯作者：
B. Sikdar
中科院分区：
文献类型：
--
作者：
Bikalpa Upadhyaya;Sumei Sun;B. Sikdar
干扰或非法无线网络访问通过模仿合法传输来干扰合法通信会话并降低网络性能。在这篇文章中，我们提出了一种方法来检测这样的攻击，通过实现一个多假设顺序测试为基础的检测框架与方差和信道状态信息（CSI）为基础的算法。检测框架的重点是区分合法和非法传输和非法传输的性质与四元假设检验。四元假设包括无传输、合法节点传输、非法节点传输和基于冲突的攻击。我们首先设计了一个三元假设问题的序贯检验问题，然后用基于方差的方法和基于CSI的方法来解决剩余的假设。我们设计算法的基础上相同的，并比较它们的性能。我们还比较了我们的方法与广义Neyman-Pearson方法的检测速度的基础上。此外，我们提出了一个多传感器为基础的方法，以进一步提高检测性能，通过软，硬决策相结合。我们根据模拟和测量数据进行广泛的性能评估。数值结果表明，所提出的算法的样本量要求更少，从而更快的检测。
Jamming or illegitimate wireless network access interferes with legitimate communication sessions by mimicking the legitimate transmissions and degrades the network performance. In this article, we propose a methodology to detect such attacks by implementing a multiple hypotheses sequential testing-based detection framework with variance and channel state information (CSI)-based algorithms. The detection framework focuses on distinguishing between legitimate and illegitimate transmissions and the nature of illegitimate transmissions with a quaternary hypotheses test. The quaternary hypotheses include no transmission, legitimate node transmission, illegitimate node transmission, and collision-based attack. We first devise a sequential testing problem on a ternary hypothesis problem and then tackle the remaining hypothesis with both variance-based approach and CSI-based approach. We devise algorithms based on the same and compare their performance. We also compare our approach with the generalized Neyman–Pearson approach based on detection speed. In addition, we present a multiple sensor-based approach to further improve the detection performance through soft- and hard-decision combining. We conduct extensive performance evaluations based on both simulated and measurement data. The numerical results show fewer sample size requirements for the proposed algorithms, leading to faster detection.