Enabling Trusted TEE-as-a-Service Models with Privacy Preserving Automatons

The ideal TEE service model foresees that data owners load their TEE software in an untrusted platform where the specific processing of the business application can take place inside an enclave, shielded against privileged attackers. In this situation, the data owner needs to trust the TEE hardware vendor only. However, it is very common that the enclave software is offered by a third-party that does not share its source code. Therefore, the service provider must be trusted as well. This is not acceptable when privacy requirements are stringent such as in the fintech ecosystem. In this paper, we propose PRIVATON, an approach based on a dual sandbox strategy leveraging TEE technologies with an embedded WebAssembly sandboxed run-time to compute privacy preserving computations modelled as finite state automatons with verifiable proofs of computations. With PRIVATON, the data owner will have the guarantee that even a malicious TEE developer will not be able to get access to the sensitive data. An implementation of PRIVATON was evaluated in the case study provided by the Credora company who is playing the role of a distributed and privacy-preserving credit oracle in the ecosystem of cryptocurrencies trading.