Oracle Extractable Value
Oracles serving Maker, Compound, and Aave on Ethereum mainnet alone forfeited a lower bounds estimate of $15,800,000 in MEV from May 21st to June 21st (calculations at the end). These same oracles likely needed to be subsidised ~ 90% due to gas costs.
The value was forfeited because oracles send every transaction that updates the data feed directly to block producers irrespective of profit opportunities the new state may create. If a new on-chain price triggers a liquidation, then bots will compete in a bidding war for the right to have their transaction ordered as soon as possible after the oracles. Block producers are the big winners currently, maintaining all the control resulting in searchers bribing them up to 95% of the profits.
Oracles have power to extract MEV by bundling updates of the data feed with liquidations either through a Flashbots-like protocol or a contract that does both transactions atomically. This can direct the value from block producers to not only the oracle, but also to dapps and users. A way for oracles to participate could be through offering “right to first liquidate auctions”, a tweak on a term coined by Hasu and something that Hayden Adams said he almost built into Uniswap V2. Ensuring separation of the oracle and MEV roles helps avoid vertical integration, while regular public auctions increase competition and transparency, both helping to mitigate the centralising forces of MEV. This monetization strategy is similar to how major tradfi exchanges operate by selling colocation services, where traders buy space right next to the exchange computers to guarantee they can receive and send orders faster than the competition.
Airnode allows for data providers to post signed data to an HTTP endpoint that third parties use to update the on-chain price in a tamper-proof way. The aim of this paper is to describe a service on top of data feeds that uses third party updates to accrue MEV back to the oracle. This service must not compromise the security of the underlying data feed, and doesn’t require data providers to have any MEV knowledge. The DAO operating the data feeds can regularly auction off access to a separate Airnode instance, or a proxy API to the Airnode (to limit the ability for anyone to compromise the original Airnode instance that powers the data feed). Auction winners can then use this access to update the on-chain price with signed data whenever there are profitable liquidation opportunities, without being beholden to the price deviation intervals that determine when the feeds are usually updated. Profit splits between searchers/dapps/oracles can be determined prior to the auction by bidders referencing an immutable and verified liquidation contract that can be audited, and then whitelisted to update the data feed. The DAO should not only be considering profits splits, but also the competency of the bidders.
Many of the issues that come with operating first party data feeds are with components like nodes/RPCs, because the data providers do not necessarily possess technical blockchain expertise. This solution incentivizes actors to backup these important components in times when protocols need the feed most, and to provide a higher update frequency. They will not only be incentivised to update the feed for the instant profits, but also in case of provider issues to protect future revenue and improve chances of winning further proposals. This is not to say that this would be more secure overall than current first party data feeds, as there may be trade-offs that come with giving more actors access to the data.
Major protocols are beginning to take advantage of abilities they have to direct MEV from block producers back to their stakeholders. An example of this is Optimism, who have been researching MEVA, which will accrue MEV to the base-layer in an auction process. Another example is KIP-17 on Rook Protocol that will allow keepers to use early access to Chainlink updates to perform liquidations and route MEV back to GMX.
“This is kind of the beginning of the cambrian explosion of MEV projects”
Phil Daian
References
https://github.com/api3dao/
https://forum.rook.fi/t/kip-17-provide-a-coordination-facility-for-gmx/272
Calculations (these are quite rough but should serve as a decent lower bounds)
Source: https://dune.com/dsalv/Liquidations on 06/21/22
I took total debt covered in the last month and multiplied that by 5% (lower bounds incentive for liquidation) to get the total profits extracted. This number assumes oracles do not need to bribe block producers.