Ultra-Low Power, Secure IoT Platform for Predicting Cardiovascular Diseases

作者：Yasin Muhammad^*; Tekeste Temesghen; Saleh Hani; Mohammad Baker; Sinanoglu Ozgur; Ismail Mohammed

来源：IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS, 2017, 64(9): 2624-2637.

DOI：10.1109/TCSI.2017.2694968

• ECG
• biomedical classifier
• ventricular arrythmia
• Internet of Things
• hardware security
• design-for-trust

Internet of Things (IoT) promises to revolutionize the health-care sector through remote, continuous, and non-invasive monitoring of patients. However, there are two main challenges faced by the IoT-enabled medical devices: energyefficiency and security/privacy concerns. Researchers have independently attempted to develop solutions, such as low-power ECG-processors and security protocols, that address these challenges on an individual basis. However, it is imperative to investigate holistic solutions that integrate in a synergistic manner, delivering an overall secure and energy-efficient product. In this paper, we develop an ultra-low power and secure IoT sensing/preprocessing platform for prediction of ventricular arrhythmia using ECG signals. Our proposed solution is able to predict the on-set of the critical cardiovascular events upto 3 h in advance with 86% accuracy. Moreover, the proposed architecture is designed using an Application Specific Integrated Circuits design flow in 65-nm Low Power Enhanced technology; the power it consumes is 62.2% less than that of the state-of-the-art approaches, while occupying 16.0% smaller area. The proposed processor makes use of ECG signals to extract a chip-specific ECG key that enables protection of communication channel. By integrating the ECG key with an existing design-for-trust solution, the proposed platform offers protection also at the hardware level, thwarting hardware security threats, such as reverse engineering and counterfeiting. Through efficient sharing of on- chip resources, the overhead of the multi-layered security infrastructure is kept at 9.5% for area and 0.7% for power with no impact on the speed of the design.

• 出版日期2017-9