Computation Oracles

So far, we have only discussed oracles in the context of requesting and delivering data. However, oracles can also be used to perform arbitrary computation, a function that can be especially useful given Ethereum’s inherent block gas limit and comparatively expensive computation costs. Rather than just relaying the results of a query, computation oracles can be used to perform computation on a set of inputs and return a calculated result that may have been infeasible to calculate on-chain. For example, one might use a computation oracle to perform a computationally intensive regression calculation in order to estimate the yield of a bond contract.

If you are willing to trust a centralized but auditable service, you can go again to Oraclize. They provide a service that allows decentralized applications to request the output of a computation performed in a sandboxed AWS virtual machine. The AWS instance creates an executable container from a user-configured Dockerfile packed in an archive that is uploaded to the Inter-Planetary File System (IPFS; see [data_storage_sec]). On request, Oraclize retrieves this archive using its hash and then initializes and executes the Docker container on AWS, passing any arguments that are provided to the application as environment variables. The containerized application performs the calculation, subject to a time constraint, and writes the result to standard output, where it can be retrieved by Oraclize and returned to the decentralized application. Oraclize currently offers this service on an auditable t2.micro AWS instance, so if the computation is of some nontrivial value, it is possible to check that the correct Docker container was executed. Nonetheless, this is not a truly decentralized solution.

The concept of a ‘cryptlet’ as a standard for verifiable oracle truths has been formalized as part of Microsoft’s wider ESC Framework. Cryptlets execute within an encrypted capsule that abstracts away the infrastructure, such as I/O, and has the CryptoDelegate attached so incoming and outgoing messages are signed, validated, and proven automatically. Cryptlets support distributed transactions so that contract logic can take on complex multistep, multiblockchain, and external system transactions in an ACID manner. This allows developers to create portable, isolated, and private resolutions of the truth for use in smart contracts. Cryptlets follow the format shown here:

public class SampleContractCryptlet : Cryptlet
  {
        public SampleContractCryptlet(Guid id, Guid bindingId, string name,
            string address, IContainerServices hostContainer, bool contract)
            : base(id, bindingId, name, address, hostContainer, contract)
        {
            MessageApi = new CryptletMessageApi(GetType().FullName,
                new SampleContractConstructor())

For a more decentralized solution, we can turn to TrueBit, which offers a solution for scalable and verifiable off-chain computation. They use a system of solvers and verifiers who are incentivized to perform computations and verification of those computations, respectively. Should a solution be challenged, an iterative verification process on subsets of the computation is performed on-chain—a kind of ‘verification game’. The game proceeds through a series of rounds, each recursively checking a smaller and smaller subset of the computation. The game eventually reaches a final round, where the challenge is sufficiently trivial such that the judges—Ethereum miners—can make a final ruling on whether the challenge was met, on-chain. In effect, TrueBit is an implementation of a computation market, allowing decentralized applications to pay for verifiable computation to be performed outside of the network, but relying on Ethereum to enforce the rules of the verification game. In theory, this enables trustless smart contracts to securely perform any computation task.

A broad range of applications exist for systems like TrueBit, ranging from machine learning to verification of proof of work. An example of the latter is the Doge–Ethereum bridge, which uses TrueBit to verify Dogecoin’s proof of work (Scrypt), which is a memory-hard and computationally intensive function that cannot be computed within the Ethereum block gas limit. By performing this verification on TrueBit, it has been possible to securely verify Dogecoin transactions within a smart contract on Ethereum’s Rinkeby testnet.