Skip to main content
DeFi typically relies on either m-of-n price feeds or AMM-derived price oracles. openOracle offers a different trust model. Compared with m-of-n feeds, openOracle does not depend on a whitelisted or KYC’d set of reporters. Its security comes from permissionless onchain participants acting in their own economic self-interest. In m-of-n systems, a central challenge is ensuring that the cost of corrupting the reporting set exceeds the profit available from corruption, especially when the value secured becomes very large. In openOracle, this security budget is parametrizable at the instance level. Compared with AMM price oracles, openOracle avoids the core limitation that manipulation losses are often bounded by AMM liquidity and fees. As Mackinga, Nadahalli, and Wattenhofer discuss in TWAP Oracle Attacks: Easier Done Than Said?, AMM-based TWAPs become vulnerable under multi-block proposer control. This problem becomes more severe in a world where proposer slots are known in advance, proposer concentration over future windows is statistical, and mechanisms like ePBS may fill in missing slots. Even median-based AMM oracles fail once an attacker controls enough proposers inside the observation window, with the attacker’s losses capped by AMM pool fees. openOracle changes the attack surface. In an AMM-derived oracle, a bad trade can poison the feed because the manipulated price becomes part of the observation history. In openOracle, a dispute fully replaces the price state with a new report, larger backing, and a reset timer. To make a false report survive, an attacker must prevent any valid dispute from landing across the entire settlement window. For a properly sized oracle game, this can be much more expensive than manipulating an AMM oracle under high proposer concentration. See Attack Vectors: Inclusion for our treatment of these attacks. In short, openOracle lets applications size oracle security per instance against the value at risk. m-of-n feeds and AMM price oracles generally cannot provide the same kind of permissionless, per-instance security scaling.