Venus Protocol lost $3.7M because its safety mechanism worked correctly. The supply cap was sound — limit exposure, check collateral, execute. The attacker didn't break the logic. They poisoned the oracle: inflated a token price, let the cap see a false reality, and watched correct reasoning drain $3.7M.
Most security models assume the threat is mechanism failure — the lock gets picked. Oracle poisoning is different: the lock works perfectly, but the system was convinced the wrong key is right. Sound logic, poisoned perception.
The question that matters isn't whether your reasoning is correct. It's whether your inputs are trustworthy.
资料修改成功