Séminaire ICE “Physically Unclonable Functions: A Device Fingerprint to Increase Security in Digital Systems”

Feb. 8th 2023, 2 pm, Lecture hall 4 and online
Physically Unclonable Functions (abridged as PUFs) are electronic structures innate to each chip instance, which generate unique per chip identifier.It becomes an inescapable security primitive as it is a native tamper-resistance identifier compared to an identification number stored by a controlled human process in a non-volatile memory. 
This talk starts with an overview about the principles of the PUF, the way the function is extracted from digital circuits, the PUF types and the main use cases for security enhancement.Then the ways to construct PUF are covered in order to efficiently use it in current and future applications. The PUF requires notably specific architectures and postprocessing to meet important properties of reliability, entropy and resistance to attacks. For instance, as the PUF is very sensitive to noise, we will see how to reduce the noise impact to get a steady identifier like a cryptographic key. This will introduce a specific public word, the “helper data”, which is necessarily associated to the PUF in order to get a reliable output. The PUF can also be attacked by different means, either physical or mathematical like the modeling attack. This latter extracts the PUF model by machine learning and allows the attacker to predict any PUF response. Hence a panel of dedicated protections will be presented to be robust against any threat.
Jean-Luc Danger  is a full professor at Institut Polytechnique de Paris / Télécom-Paris. His research focuses on secure hardware, randomness management, and implementation of complex / constrained algorithms. He is particularly interested in analysis and protections of embedded systems  against physical attacks,  security primitives like True Random Generator and Physically Unclonable Functions. He leads  the “Secure and Safe Hardware” group at Telecom-Paris and is the deputy head of the LTCI laboratory.