Visible to the public A Truth-Inducing Sybil Resistant Decentralized Blockchain Oracle

TitleA Truth-Inducing Sybil Resistant Decentralized Blockchain Oracle
Publication TypeConference Paper
Year of Publication2020
AuthorsCai, Y., Fragkos, G., Tsiropoulou, E. E., Veneris, A.
Conference Name2020 2nd Conference on Blockchain Research Applications for Innovative Networks and Services (BRAINS)
Keywordscomposability, cryptography, decentralized oracle protocols, Decentralized Oracles, Distributed databases, majority-voting schemes, maximized expected score, Metrics, nonlinear stake scaling rule, Peer Prediction, peer prediction scoring scheme, pubcrawl, Resiliency, Staked Voting, sybil attacks, truth-inducing Sybil resistant decentralized blockchain oracle
AbstractMany blockchain applications use decentralized oracles to trustlessly retrieve external information as those platforms are agnostic to real-world information. Some existing decentralized oracle protocols make use of majority-voting schemes to determine the outcomes and/or rewards to participants. In these cases, the awards (or penalties) grow linearly to the participant stakes, therefore voters are indifferent between voting through a single or multiple identities. Furthermore, the voters receive a reward only when they agree with the majority outcome, a tactic that may lead to herd behavior. This paper proposes an oracle protocol based on peer prediction mechanisms with non-linear staking rules. In the proposed approach, instead of being rewarded when agreeing with a majority outcome, a voter receives awards when their report achieves a relatively high score based on a peer prediction scoring scheme. The scoring scheme is designed to be incentive compatible so that the maximized expected score is achieved only with honest reporting. A non-linear stake scaling rule is proposed to discourage Sybil attacks. This paper also provides a theoretical analysis and guidelines for implementation as reference.
DOI10.1109/BRAINS49436.2020.9223272
Citation Keycai_truth-inducing_2020