CAD for Assurance of Electronic Systems

Deep Learning Based Model Building Attacks on Arbiter PUF Compositions

By: Pranesh Santikellur (IIT Kharagpur), Aritra Bhattacharyay (University of Florida), Rajat Subhra Chakraborty (IIT Kharagpur)

Stage: On Field (Runs on Basic Workstation)


The tool provides access to deep learning based model building attacks on various arbiter PUF (APUF) based compositions. It covers nearly twenty different PUF designs, making it comprehensive collection on model building attacks. The deep learning based attack was implemented in Python 2.7 using Keras and Tensorflow libraries. The dataset for each PUF is also provided for easy reproduction of attacks. For more details, please refer the link:


Pranesh Santikellur

Input/Output Interface

  • Input: Dataset consisting of parity vectors and responses of the corresponding PUF
  • Output: Accuracy of the modeling attack


  • Python 2.7
  • Tensorflow
  • Keras
  • scikit-learn


Santikellur, Pranesh; Bhattacharyay, Aritra; Chakraborty, Rajat Subhra

Deep Learning based Model Building Attacks on Arbiter PUF Compositions Miscellaneous

Cryptology ePrint Archive, Report 2019/566, 2019.

Abstract | Links | BibTeX

User Guide Link can be found at this link