Deep Learning based Security Enhancement of Wireless Connectivity

Implementing and strengthening wireless connectivity security functions for sensor nodes, including active security concepts such as data acquisition authentication of IoT sensor node devices and control authority authentication related to device control, is critical for protecting people’s lives, the economy and functions of society, and security. Lightweight cryptographic methods are required for resource-constrained sensor node devices such as smart meter and implantable medical device (IMD). Physical unclonable function (PUF) is a promising hardware-based lightweight security solution that generates a unique ID by utilizing inherent process variation in the semiconductor fabrication process. In this work a simple authentication protocol with PUF for lightweight operation is presented. In addition, a low-power, lightweight analog inference circuit-based deep learning function is proposed for enhanced wireless connectivity security.