Abstract
There has been an unprecedented surge in the number of service providersoffering a wide range of machine learning prediction APIs for tasks such asimage classification, language translation, etc. thereby monetizing theunderlying data and trained models. Typically, a data owner (API provider)develops a model, often over proprietary data, and leverages the infrastructureservices of a cloud vendor for hosting and serving API requests. Clearly, thismodel assumes complete trust between the API Provider and cloud vendor. On theother hand, a malicious/buggy cloud vendor may copy the APIs and offer anidentical service, under-report model usage metrics, or unfairly discriminatebetween different API providers by offering them a nominal share of therevenue. In this work, we present the design of a blockchain baseddecentralized trustless API marketplace that enables all the stakeholders inthe API ecosystem to audit the behavior of the parties without having to trusta single centralized entity. In particular, our system divides an AI model intomultiple pieces and deploys them among multiple cloud vendors who thencollaboratively execute the APIs. Our design ensures that cloud vendors cannotcollude with each other to steal the combined model, while individual cloudvendors and clients cannot repudiate their input or model executions.