Current Mirror Array: A Novel Circuit Topology for Combining Physical Unclonable Function and Machine Learning

Edge analytics support industrial Internet of Things by pushing some data processing capacity to the edge of the network instead of sending the streaming data captured by the sensor nodes directly to the cloud. It is advantageous to endow machine learners for data reduction with suitable security primitives for privacy protection in edge computing devices to conserve area and power consumption. In this paper, we propose a novel physical unclonable function (PUF) based on current mirror array (CMA) circuits that reuses the circuit implementation of a machine learner-the extreme learning machine (ELM), which is a randomized neural network. Seven different challenge activation and response readout schemes are proposed to realize different weak and strong PUF functions from within the same CMA array. ELM endowed with such reconfigurable challenge-response mechanism is more robust and adaptable to different authentication protocols and security functions. Measurement results on 0.35 mu m test chips demonstrate that the proposed strong PUF outperforms other state-of-the-art designs with smaller area/bit of 9 x 10(-36) mu m(2) and lower native bit error rate (BER) of 0.16% with an added overhead of less than 2.5% power and 2.9% area over the native ELM implementation.