ERI: EMRadar: A Practical Sensing System on the Electromagnetic Side-Channel of IoT

物联网（IoT）已经彻底改变了人们的日常生活，但由于物联网设备上的资源有限，以及一旦部署后对其内部操作缺乏可视性，确保物联网设备的安全仍然是一项重大挑战。与传统的侧信道漏洞不同，电磁（EM）侧信道发射通过无线信道泄露信息，提供了一种独特的媒介，可以在不访问或检测设备的情况下调查物联网安全性。不幸的是，目前对物联网设备的电磁侧信道攻击的研究受到了严重的限制，因为缺乏一种实用的传感系统，可以提取计算机处理器或存储器发出的弱电磁信号，分离共存设备的这些电磁信号，并强大地揭示其语义。为了应对这些挑战，该项目将设计、实施和评估一个名为EMRadar的实用电磁传感系统，以研究真实环境中的电磁侧信道攻击和防御。该项目的结果将加深我们对电磁侧信道攻击的理解，并为排放安全标准提供信息。该项目实现的实用EM侧通道防御将改变各种安全关键型物联网应用，包括检测资源高度受限的医疗设备（如植入式心脏设备和智能胰岛素泵）中的程序偏差。该项目致力于通过将研究与教育和外展活动相结合，促进STEM领域的多样性和包容性。PI将通过讲解物联网技术和网络安全基本概念的实践活动和演讲，积极吸引当地高中学生。为了弥合STEM参与方面的差距，该项目将吸引来自底特律和庞蒂亚克未被充分代表的少数群体的学生，以及密歇根州的各种女孩组织。这项努力得到了与奥克兰大学STEM夏令营和实地考察以及密歇根计算机委员会的合作的支持。通过这些举措，该项目将激发广泛的学生追求工程和科学的未来。该项目将设计、实施和评估EMRadar，这是一种实用的电磁传感系统，可以在现实环境中研究电磁侧信道攻击和防御。提出的研究将建立在一个三层架构模型上。(1)在传感层，EMRadar将采用创新的信号处理技术，提取深埋在噪声中、受干扰污染的微弱电磁信号。为了实现这一目标，EMRadar将利用处理器或存储器和物联网软件行为的电磁发射的独特特性，实现新的方法，可以适应和优化经典雷达系统中使用的信号处理技术，以实现高灵敏度的电磁传感。(2)在表示层，本项目将探索EM信号的通用表示和分类，以揭示处理器和存储器的EM侧通道发射。尽管对电磁侧通道发射进行了广泛的研究，但由于电磁信号模式的显著变化以及在同一设备上运行的工作负载产生的电磁干扰，对电磁侧通道信号语义的准确解释仍然极具挑战性。该项目将利用时间序列分析和机器学习的最新进展来重新审视EM信号的表示，并进行系统的测量研究，以了解具有不同架构的物联网设备的通用表示的鲁棒性。(3)在应用层，EMRadar将在实际设置和应用中进行评估。PI将与领域专家合作，探索该项目开发的系统的应用，例如在医疗保健、汽车工业和智能建筑中支持高效的基于物联网的应用。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The Internet of Things (IoT) has revolutionized daily life, yet securing IoT devices remains a significant challenge due to limited resources on IoT devices and the lack of visibility into their internal operations once deployed. Unlike conventional side-channel vulnerabilities, the electromagnetic (EM) side-channel emission leaks information over wireless channel, offering a unique medium to investigate IoT security without accessing or instrumenting the device. Unfortunately, current research on the EM side-channel attacks of IoT devices has been severely limited by the lack of a practical sensing system that can extract weak EM signals emitted by computer processors or memories, separate those EM signals of co-located devices, and robustly unravel their semantics. To tackle these challenges, this project will design, implement, and evaluate a practical EM sensing system named EMRadar to enable research on EM side-channel attacks and defenses in real environments. The results of this project will deepen our understanding of EM side-channel attacks and inform emission security standards. Practical EM side-channel defenses enabled by this project will transform various security-critical IoT applications, including detecting program deviations in medical devices that are highly resource-constrained, such as the implantable cardiac devices and the smart insulin pumps. The project is dedicated to fostering diversity and inclusion in STEM by blending research with education and outreach activities. The PI will actively involve local high school students by providing hands-on activities and presentations elucidating fundamental concepts of IoT technologies and cybersecurity. To bridge the gap in STEM participation, the project will engage students from underrepresented minority groups in Detroit and Pontiac as well as various Michigan-based organizations for girls. This endeavor is supported by collaborations with the Oakland University's STEM summer camps and field trips, as well as the Michigan Aspirations in Computing Committee. Through these initiatives, the project will inspire a broad spectrum of students to pursue futures in engineering and science.This project will design, implement, and evaluate the EMRadar, a practical EM sensing system that enables research on EM side-channel attacks and defenses in realistic environments. The proposed research will build on a three-layer architecture model. (1) In the sensing layer, EMRadar will incorporate innovative signal processing techniques to extract weak EM signals that are deeply buried in noise and contaminated by interference. To achieve this, the EMRadar will exploit the unique characteristics of EM emissions from a processor or memory and IoT software behavior, enabling novel methods that can adapt and optimize signal processing techniques used in classic Radar systems for highly sensitive EM sensing. (2) In the representation layer, this project will explore universal representation and classification of EM signals to unravel EM side-channel emissions of processors and memories. Despite extensive studies on EM side-channel emissions, accurate explanation of the semantics of EM side-channel signals remains highly challenging due to significant variation of EM signal pattern and the EM interference produced by workloads running on the same device. This project will leverage the recent advances in time series analysis and machine learning to revisit the representation of EM signals, and conduct a systematic measurement study to understand the robustness of universal representation across IoT devices with different architectures. (3) In the application layer, the EMRadar will be evaluated in real-world setups and applications. The PI will collaborate with domain experts to explore applications of the system developed in this project, such as supporting efficient IoT-based applications in healthcare, automobile industry, and intelligent buildings.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.