Figure 4.

Strong A PUF resilient to machine learning attacks. a) Strong PUF architecture with a final 1‐bit response. Challenge consists of 32 bits, where each bit determines whether the switch is turned on to generate intermediate bits for XOR cascade. b) Strong‐PUF flowchart showing how the strong PUF is achieved. c) Conceptual diagram showing the repeated PUF configuration exhibited as an unrolled cascade of PUF without repeating processes. After each repeating step using different challenges, the process is stopped to enable the output bit R(n) to be read and inserted into the response matrix until 100 bits are generated to form the 10 × 10 key. d) Machine learning (ML) accuracy as a function of CRPs utilized for training and with two different ML models, logistic regression (LR) and multilayer perception (MLP). e,f) Inter‐HD histograms for the e) UQ and f) UF.