SAF-PUF: A Strong PUF with Zero-BER, ML-Resilience and Dynamic Key Concealment Enabled by RRAM Stuck-at-Faults

This repository hosts the supporting data and source code for the paper titled "SAF-PUF: A Strong PUF with Zero-BER, ML-Resilience and Dynamic Key Concealment Enabled by RRAM Stuck-at-Faults." Folder Structure: Folder 2.2 Response Generation and Performance MetricsCorresponds to Section 2.2 of the paper.Contains code for the SAF-PUF design and for calculating the performance metrics (Uniformity, Diffuseness, Uniqueness). Folder 2.4 Resilience to Machine Learning AttacksCorresponds to Section 2.4 of the paper.Subfolder Code and data: Contains the initial dataset for neural network training and the training code for three machine learning models: Logistic Regression (LR), Support Vector Machine (SVM), and Neural Network (NN).Subfolder Results: Contains the training and test accuracy results for the three machine learning models across datasets of varying sizes. Folder 2.5 Key concealment and Recovery MechanismCorresponds to Section 2.5 of the paper.Subfolder Code: Contains code for implementing and evaluating the "use-then-conceal" functionality.Subfolder Results: Contains data on the Identical Response Rate and the Bit-Flip Rate as the RRAM array size scales from 16×16 to 256×256.