CAREER: AI-Enabled Self-Healing and Trusted Wireless Transceivers for Biomedical Applications

在后新冠肺炎时代，医学界越来越多地采用远程医疗作为传统医学的替代方案。无线连接的生物医疗设备是此类远程医疗解决方案不可或缺的一部分。随着患者寿命的延长，用于生物医学设备的无线收发器的长期可靠性正成为一个令人担忧的问题。此外，个性化医疗保健的敏感性引发了人们对数据安全的担忧。这个职业项目将研究射频集成电路(RFIC)中的安全威胁和故障机制，并开发智能、低能耗的模拟解决方案，以创建具有检测和治愈损伤能力的可信无线收发器。这一职业项目中提出的研究将从根本上改变远程医疗解决方案，开发能够在设备级别提供高速连接的新型小型化自我修复和可信的无线收发器。此外，该项目通过引入用于安全数据通信的新型低能量模拟加密技术来增强硬件安全的基础知识。该项目的教育计划将显著提高学生在通信和硬件安全方面的知识。通过与行业的合作，学生将有机会与行业导师合作，并获得电子行业的实用知识。该计划还包括侧重于K-12阶段STEM教育的举措。作为这项努力的一部分，将向当地高中生及其教师提供关于安全电子和数据通信的暑期讲习班，随后将举行设计比赛，以激励学生，特别是那些来自代表不足的群体和少数族裔的学生，寻求与STEM相关的专上教育。作为推广活动的一部分，还将组织实验室访问和新兵训练营，以分享资源并促进向教师和学生传递知识。该职业项目的目标是通过引入低能耗模拟非对称加密和自适应自愈来开发智能、自愈和可信的无线收发器。提出的研究包括两个主要领域：(1)具有智能威胁检测能力的低能量可信数据通信；(2)智能自愈。这两项研究都得益于节能的模拟神经网络(ANN)来提高功能。对无线收发机中的调制波形应用创新的非对称模拟加密，创建了节能的端到端加密无线通信链路。所提出的设计将使用可伸缩的线性时不变(LTI)来生成加密和解密所需的密钥。低能量人工神经网络将监控收发机参数，以发现任何潜在攻击的迹象，并相应地通知加密引擎。收发器还将使用低开销设备指纹技术进行身份验证。同样，将开发一个低能量自适应模拟自愈单元，通过检测性能下降和异常来提高可靠性，并主动调整收发参数。自愈单元采用创新的双环自适应结构。第一个环路构建在模拟前端(AFE)内，始终用于监控短期性能下降，而第二个环路依赖于低能量ANN，并选择性地对数据进行采样，以纠正长期缺陷。这两个研究推力中使用的神经网络都是使用节能的尖峰神经网络(SNN)构建的，并共同设计以减少延迟和区域开销。这一职业项目的成果将通过加速在医疗界采用智能个性化解决方案来增强远程医疗，特别是在慢性病的长期治疗方面。它还将为生物医学应用奠定一个智能、自我修复和值得信赖的电子平台的基础。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

In the post COVID-19 era, the medical community is increasingly adopting remote healthcare as an alternative to conventional medicine. Wirelessly-connected biomedical devices are an indispensable part of such remote healthcare solutions. With the increased longevity of patients, the long-term reliability of the wireless transceivers used in biomedical devices is becoming a concern. Furthermore, the sensitive nature of personalized healthcare raises concerns about the data security. This CAREER project will study the security threats and failure mechanisms in radio frequency integrated circuits (RFIC) and develop intelligent, low-energy analog solutions so as to create a trusted wireless transceiver with the capability to detect and cure impairments. The proposed research in this CAREER project will fundamentally change the remote healthcare solutions by developing a new class of miniaturized self-healing and trusted wireless transceivers that can provide high-speed connectivity at the device level. Moreover, this project enhances the foundational knowledge in hardware security by introducing novel low-energy analog encryption techniques for use in secure data communications. The education plan in this project will significantly enhance the knowledge of students in communications and hardware security. Through collaboration with industry, students will have the opportunity to work with industrial mentors and gain practical knowledge in electronics. The plan also contains initiatives focused on STEM education in K-12. As part of this effort, summer workshops on secure electronics and data communications will be offered to local high school students and their teachers, followed by design competitions to inspire the students, particularly those from underrepresented groups and minorities, to seek post-secondary education in STEM related fields. Lab visits and boot camps will also be organized as part of outreach activities to share resources and facilitate the knowledge transfer to teachers and students.The goal of this CAREER project is to develop intelligent, self-healing and trusted wireless transceivers by introducing low-energy analog asymmetric encryption and adaptive self-healing. The proposed research consist of two research thrust areas: (1) low-energy trusted data communications with smart threat detection capability, and (2) intelligent self-healing. Both research thrusts benefit from energy-efficient analog neural networks (ANNs) to improve the functionality. Applying innovative asymmetric analog encryption on the modulated waveforms in the wireless transceiver, an energy-efficient end-to-end encrypted wireless communication link is created. The proposed design will use scalable linear time-invariant (LTI) to generate the keys needed for encryption and decryption. A low-energy ANN will monitor the transceiver parameters for any sign of potential attack and notify the encryption engine accordingly. The transceiver will also be authenticated using low-overhead device fingerprinting techniques. Similarly, a low-energy adaptive analog self-healing unit will be developed to increase the reliability by detecting the performance degradation and abnormalities, and actively adjust the transceiver parameters. The self-healing unit uses an innovative dual-loop adaptive structure. The first loop is built within the analog front-end (AFE) and is always engaged to monitor short-term performance degradations while the second loop relies on a low-energy ANN and samples the data selectively to correct for long-term defects. The ANNs used in both research thrusts are built using energy-efficient spiking neural networks (SNNs) and are co-designed to reduce the delay and area overhead. The outcome of this CAREER project will enhance the remote healthcare by accelerating the adoption of smart personalized solutions in medical community, particularly for long-term treatment of chronic diseases. It will also lay the foundation of a smart, self-healing and trusted electronic platform for biomedical applications.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.