Blockchains can't naturally talk to each other — Bitcoin doesn't know Ethereum exists, and your ETH can't simply travel to Solana. Bridges solve this, usually by locking your coins in a vault on chain A and minting an IOU ("wrapped") version on chain B. That design creates something irresistible to attackers: a single vault holding enormous value, guarded by complex code and, often, a handful of keys.
Why bridges specifically?
Three reasons. Concentration: a popular bridge's vault can hold hundreds of millions — the biggest honeypot in crypto. Complexity: a bridge must correctly understand events on two different chains at once; every extra moving part is attack surface. Trust shortcuts: many bridges are secured not by a blockchain's full consensus but by a small committee of validator keys — steal enough keys and the vault opens. Most mega-hacks exploited the third.
The hall of shame
Poly Network — $611M (Aug 2021). A code flaw let one attacker drain the largest sum in DeFi history. In a bizarre twist, the hacker returned essentially all of it, claiming it was "for fun" — the industry's luckiest disaster.
Wormhole — $325M (Feb 2022). A signature-verification bug on the Solana side let the attacker mint 120,000 wrapped ETH from nothing. Trading firm Jump Crypto refilled the vault from its own pocket — a private bailout that saved the bridge.
Ronin — $625M (Mar 2022). The bridge behind Axie Infinity was run by just nine validator keys, five of which could move everything. North Korea's Lazarus Group phished its way to five keys — partly via a fake job offer PDF — and quietly emptied the vault. Nobody noticed for six days. The largest crypto hack ever, and barely any "hacking" involved.
Harmony Horizon — $100M (Jun 2022). A 2-of-5 multisig. Lazarus again. Two keys were enough.
Nomad — $190M (Aug 2022). A botched update made any withdrawal message look valid. Once one attacker showed the trick, hundreds of copycats simply pasted the transaction with their own address — a first-of-its-kind crowd-looting.
Multichain — $126M+ (Jul 2023). The bridge's keys turned out to be effectively controlled by its CEO — who was arrested in China, after which funds moved out and the bridge died. Chains that depended on it (notably Fantom) suffered lasting ecosystem damage. The lesson: a "decentralised" bridge can secretly be one man with a laptop.
The patterns
Almost every bridge disaster fits one of three patterns: stolen keys (Ronin, Harmony, Multichain — social engineering or centralisation, not clever math), code bugs (Wormhole, Nomad, Poly — one flawed check in a complex machine), or rug-adjacent governance (whoever controls upgrades controls the vault). Notice what's missing: the underlying blockchains were never broken. The chains are the strong part; the connections between them are the weak part.
How to reduce your risk
Prefer native assets. Holding real ETH on Ethereum beats holding wrapped ETH on another chain — the wrapper is only as good as its bridge's vault. Use official routes. A chain's canonical bridge (or an exchange withdrawal directly to the target chain) usually carries less exotic risk than third-party bridges. Don't park value in wrappers. Bridge, do what you came to do, and avoid leaving large balances in bridged IOUs long-term. Know what backs your token. If you hold a wrapped asset, you're a creditor of a bridge — find out which one. And treat "audited" as necessary, not sufficient: most of the bridges above had audits.
Is it getting better?
Slowly, yes. Newer designs (light-client and zero-knowledge verification) replace trusted committees with math, letting one chain cryptographically verify another's state. Ethereum's Layer-2 rollups route around the problem for their ecosystem with native bridges secured by Ethereum itself. And messaging standards like LayerZero and Axelar compete on security models, not just speed. But wherever a big trusted vault still exists, assume Lazarus is already studying it.