Reinforcement Learning Toward Decision-Making for Multiple Trusted-Third-Parties in PUF-Cash

Fragkos, Georgios; Minwalla, Cyrus; Plusquellic, Jim; Tsiropoulou, Eirini Eleni

doi:10.1109/wf-iot48130.2020.9221344

Cited by 6 publications

(1 citation statement)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Alice’s discrete action space consists of vectors

, where

, and

where

is the total number of subsets of the

nodes with cardinality

[ 22 ]. Alice aims at minimizing her communication delay with the selected nodes, thus, she prefers to select nodes with small physical distance and low congestion

[ 23 , 24 ]. Moreover, she considers the nodes’ scores, as they are reported by them by transmitting respective wireless beacons ( Section 2.3 ), while also weighing the reported values based on her probabilistic trust belief.…”

Section: Autonomous Reinforcement Learning-based Contributors Selectionmentioning

confidence: 99%

Local Trust in Internet of Things Based on Contract Theory

Fragkos

Minwalla

Plusquellic

et al. 2022

Sensors

Self Cite

View full text Add to dashboard Cite

Autonomous trust mechanisms enable Internet of Things (IoT) devices to function cooperatively in a wide range of ecosystems, from vehicle-to-vehicle communications to mesh sensor networks. A common property desired in such networks is a mechanism to construct a secure, authenticated channel between any two participating nodes to share sensitive information, nominally a challenging proposition for a large, heterogeneous network where node participation is constantly in flux. This work explores a contract-theoretic framework that exploits the principles of network economics to crowd-source trust between two arbitrary nodes based on the efforts of their neighbors. Each node in the network possesses a trust score, which is updated based on useful effort contributed to the authentication step. The scheme functions autonomously on locally adjacent nodes and is proven to converge onto an optimal solution based on the available nodes and their trust scores. Core building blocks include the use of Stochastic Learning Automata to select the participating nodes based on network and social metrics, and the formulation of a Bayesian trust belief distribution from the past behavior of the selected nodes. An effort-reward model incentivizes selected nodes to accurately report their trust scores and contribute their effort to the authentication process. Detailed numerical results obtained via simulation highlight the proposed framework’s efficacy and performance. The performance achieved near-optimal results despite incomplete information regarding the IoT nodes’ trust scores and the presence of malicious or misbehaving nodes. Comparison metrics demonstrate that the proposed approach maximized the overall social welfare and achieved better performance compared to the state of the art in the domain.

show abstract