Organizer: CASED, CROSSING and EC SPRIDE
The general concept of Physically Unclonable Functions (PUFs) has been nowadays widely accepted and adopted to meet the requirements of secure identification and key generation/storage for cryptographic ciphers.
However, shattered by different kinds of attacks, it has been earlier proved that the promised security features of arbiter PUFs, including unclonability and unpredictability do not hold unconditionally. In this talk we will present some of our recent breakthroughs in understanding the limits of the “real” security as offered by the arbiter PUF family. We will outline a Complete and Linear Physical Characterization Methodology for the arbiter PUF Family and as well a Probably Approximately Correct (PAC) learning algorithm for the Arbiter PUF Family.
Our proof of the PAC learnability of arbiter PUFs reveals many new insights into the correct construction of “really” secure arbiter PUFs in general, while our physical characterization itself proves that arbiter PUF do not deserve the name PUF at all.