Oracle Manipulation

仕組み

Contracts cannot see outside their own chain, so they rely on oracles: an on-chain automated market maker's spot price, a median of exchange quotes pushed on-chain, or a signed price pulled at transaction time. If the venue behind the feed has thin depth relative to the value the protocol will lend or settle against it, an attacker can buy or sell into that venue, move the reported price, and act on the distorted value inside the same block or the same few minutes. Time-weighted averages raise the cost by requiring the price to be held away from fair value for longer, but they also make the feed stale during genuine fast moves. Failure modes run both ways: a manipulated price can inflate collateral and enable borrowing that is never repaid, and a stale price can prevent liquidations that should have happened.

実際に観測できるもの

Read the oracle configuration: how many independent sources feed it, whether it is a spot read or a time-weighted average and over what window, what deviation and heartbeat thresholds trigger an update, and what the contract does when the feed goes stale. Compare the depth of the underlying market for a collateral asset against the maximum amount borrowable against it, since manipulation cost scales with that depth. Governance records often show when a market was listed with parameters set before its liquidity existed.

先例

In 2022 a trader pushed the price of a thinly traded perpetual contract on Mango Markets far above prevailing levels, used the inflated collateral value to borrow against it, and left the protocol with unpayable debt.

重要性を左右する要因

Check the depth of the priced market relative to the value that depends on it, the number of independent sources, the averaging window, whether a circuit breaker exists, and whether governance can list new collateral without a delay.