Sourcing Trust From Peers with Physical Unclonable Functions

Traditional authentication schemes rely heavily on trusted third parties that act as certificate authorities to source authentication credentials. In this paper, we describe an offline, end-to-end mutual authentication and session key generation protocol, called PUF-based Peer Trust (PPT), where authentication and session key generation between two parties (Alice and Bob) is implemented using physical unclonable functions (PUFs) and authentication information from neighboring peer devices (Charlies). Trust among the parties is established by defining an effort-reward model based on Contract Theory that leverages the trust beliefs provided by a set of peer devices. Alice autonomously selects a set of trusted Teds from the set of untrusted Charlies using a reinforcement learning algorithm based on Stochastic Learning Automata. These Teds participate in a peer trust authentication process designed to help Alice decide if Bob is trustworthy. The Teds are incentivized to provide accurate information (effort) via a reward system based on trust scores. Alice computes a personalized trust belief distribution for all the Charlies and utilizes it to eventually update the trust scores of the selected Teds based on their contribution to Bob's authentication process. An experimental evaluation of the model is carried out using a set of PUF-instantiated FPGAs. Detailed numerical results are presented demonstrating the effectiveness of the proposed scheme and its ability to correctly decide if Bob is trustworthy based on incomplete information provided by peers.