Orthos: A Trustworthy AI Framework for Data Acquisition

Abstract

Information acquisition through crowdsensing with mobile agents is a popular way to collect data, especially in the context of smart cities where the deployment of dedicated data collectors is expensive and ineffective. It requires efficient information elicitation mechanisms to guarantee that the collected data are accurately acquired and reported. Such mechanisms can be implemented via smart contracts on blockchain to enable privacy and trust. In this work we develop Orthos, a blockchain-based trustworthy framework for spontaneous location-based crowdsensing queries without assuming any prior knowledge about them. We employ game-theoretic mechanisms to incentivize agents to report truthfully and ensure that the information is collected at the desired location while ensuring the privacy of the agents. We identify six necessary characteristics for information elicitation mechanisms to be applicable in spontaneous location-based settings and implement an existing state-of-the-art mechanism using smart contracts. Additionally, as location information is exogenous to these mechanisms, we design the Proof-of-Location protocol to ensure that agents gather the data at the desired locations. We examine the performance of Orthos on Rinkeby Ethereum testnet and conduct experiments with live audience.