Oracle Manipulation Attacks

Many DeFi protocols — lending platforms, derivatives venues, synthetic-asset systems — need a live price for an asset to decide how much collateral is enough, or what a trade should cost. An oracle is the mechanism that supplies that price on-chain. An oracle manipulation attack exploits a protocol that trusts a price source an attacker can temporarily move, rather than one that's expensive or slow to distort.

The classic version targets a protocol that reads its price directly from a single decentralized exchange pool's spot price. Because a pool's spot price is just whatever ratio its reserves currently sit at, a large enough trade shifts it — and with a flash loan, an attacker can borrow a huge, uncollateralized sum for a single transaction, push the pool's price far from its true market value, exploit a second protocol that's reading that distorted price (say, borrowing far more than real collateral value against an inflated valuation), and repay the flash loan, all within one atomic transaction, before the price reverts.

For example: an attacker flash-borrows a large amount of token A, dumps it into a thin liquidity pool to crash token A's on-chain price, then uses token A as collateral (now valued by the manipulated feed at a stale, higher price) to borrow far more of token B than the real collateral justifies from a lending protocol that trusts that same pool's price — walking away with the difference once both loans are settled in the same transaction.

The fix is architectural rather than reactive: use time-weighted average prices instead of instantaneous spot prices (so a single large trade can't move the feed much within one block), pull from multiple independent, deep-liquidity sources and require them to agree, or use purpose-built oracle networks that aggregate off-chain data with economic security of their own. A protocol's exposure to this attack is really a question of how cheap it is to move whatever price source it trusts, even briefly.