IoT Sentinel: Correlation-based Attack Detection, Localization, and Authentication in IoT Networks
IoT Sentinel: Correlation-based Attack Detection, Localization, and Authentication in IoT Networks
复制标题
DOI：
10.1109/icccn58024.2023.10230209
复制
发表时间：
2023-07
期刊：
影响因子：
--
通讯作者：
Dianshi Yang;Abhinav Kumar;Stuart Ray;Wei Wang;R. Tourani
中科院分区：
文献类型：
--
作者：
Dianshi Yang;Abhinav Kumar;Stuart Ray;Wei Wang;R. Tourani
安全问题已成为物联网（IoT）网络面临的主要挑战之一。为了克服这一挑战，近期常用的方法主要集中在对物联网通信进行加密，或通过使用预共享凭证（例如密码和无线信道特征）对物联网设备进行持续认证。然而，这些机制被认为是不够的，部分原因是数据泄露事件日益增多，以及近期敏感的物联网设备和应用大量涌现。我们提出了物联网哨兵（IoT Sentinel）——一种新颖的安全系统，它探索物联网设备之间的相关性，以高效且有效地保护物联网网络。具体而言，我们的系统（i）检测潜在攻击，（ii）定位攻击者，以及（iii）同时进行动态隐式认证。此外，物联网哨兵不需要完全访问物联网设备的物理层来对无线信号进行精细测量，它仅使用粗粒度的数据包级设备相关性信息来保护物联网网络，对网络造成的开销可忽略不计。因此，我们的方法与现有的受限物联网设备兼容。我们在不同的场景和设置下广泛评估了物联网哨兵的功效。实验结果表明，我们的方法实现了约96%的攻击检测准确率、超过70%的攻击者定位准确率以及约100%的设备认证准确率。
Security issues have become one of the major challenges for Internet-of-Things (IoT) networks. To overcome this challenge, the recent commonly-used approaches mainly focus on conducting encryption on IoT communication or performing continuous authentication for IoT devices by using pre-shared credentials (e.g., passcode and wireless channel signatures). However, these mechanisms are deemed insufficient, in part, due to the increasing number of data breaches and the recent proliferation of sensitive IoT devices and applications. We present IoT Sentinel - a novel security system that explores the correlation between IoT devices to effectively and efficiently secure IoT networks. Specifically, our system (i) detects potential attacks, (ii) localizes the attacker, and (iii) conducts dynamic implicit authentication at the same time. Moreover, instead of requiring full physical-layer access to IoT devices for finegrained measurement of the wireless signal, IoT Sentinel uses only coarse packet-level device correlation information to secure IoT networks with negligible overhead to the network. Thus, making our approach compatible with existing constrained IoT devices. We extensively evaluate the efficacy of IoT Sentinel in different scenarios and settings. The experiment results show that our approach achieves around 96% attack detection accuracy, more than 70% attacker localization accuracy, and around 100% device authentication accuracy.