Physical Unclonable Functions: Machine Learning and Security Challenges

Explore the intricacies of Physical Unclonable Functions (PUFs) in this comprehensive conference talk. Delve into machine learning applications for PUFs, addressing the challenges of simulation inconsistencies and practical implementations. Examine the combination of optimization objectives and the concept of strong paths in PUF design. Analyze various attack models, including modeling attacks and reliability attacks, while considering the technical framework and attack complexity. Investigate the role of symmetry, Pseudorandom Functions (PRFs), and masking techniques in enhancing PUF security. Gain insights into learning with parity noise and understand the significant contributions made in this field. Conclude with a summary and engage in a Q&A session to further explore the physical aspects of cryptographic systems.

Introduction
Machine Learning of Physical Unclonable Functions
Inconsistency of Simulation and Practice
Combining Optimization Objectives
Strong Path
Modeling Attacks
Reliability Attacks
Tech Framework
Attack Complexity
Symmetry
PRF
Contributions
Summary
Masking
Questions
Learning with parity noise