Cloning Optical Physical Unclonable Functions Using Machine Learning Attacks


Physical Unclonable Functions (PUFs) can be used to uniquely identify instances of physical objects that are otherwise constructed in the same way. Several different types of PUFs exist including optical PUFs. Current developments of the optical PUF include the integrated optical PUF which is practically feasible for the integration into microelectronic systems. In this work, the integrated optical PUF was implemented and attacked. Our research confirmed that attacks using machine learning on the integrated optical PUF are possible.