Back to: Smart Contracts for Beginners
Theory ends here. Each of these hacks taught the industry a rule — learn them cheaply, from history instead of your wallet.
The DAO (2016) — reentrancy
An attacker recursively withdrew ETH before the balance updated — $60M gone, Ethereum forked to undo it. Rule: update state *before* sending funds (checks-effects-interactions), and get audited.
Parity multisig (2017) — unprotected init
Someone claimed ownership of a shared library contract and self-destructed it, freezing ~$150M permanently. Rule: initialization functions need access control, and shared libraries are single points of failure.
Oracle exploits (2020–forever) — price lies
Attackers used flash loans to warp thin-pool prices that lending protocols trusted, borrowing millions against fake collateral value. Rule: oracles must be manipulation-resistant (TWAPs, multiple sources, circuit breakers).
Bridge hacks (2022–2023) — trusted verifiers
Ronin ($625M): compromised validator keys. Wormhole ($325M): signature verification bug. Nomad ($190M): a routine upgrade zeroed the security check and *anyone* could copy-paste the exploit. Rule: bridges concentrate risk exactly where attackers look — verify, multisig, and bug bounties or don’t bridge size.
The pattern across all of them
Small code, huge assumptions: “callers are honest,” “prices are true,” “keys are safe,” “upgrades are careful.” Audits catch bug *classes*, not broken assumptions — which is why economic design review matters as much as code review.
Every hack was someone’s tuition. The curriculum is public, the tuition is
> optional — study the case files before depositing anywhere new.
Next lesson: oracles and off-chain data.