The Ronin Bridge Hack: Anatomy of a 600 Million Dollar Loss
About 625 million dollars stolen. Six full days before anyone noticed. Five of nine validator nodes compromised.
Those three numbers more or less summarize the Ronin bridge hack of March 2022, at the time the largest single crypto theft outside of Mt.Gox. The attackers didn’t ship a clever new smart contract exploit. They didn’t break any cryptography. They simply convinced five of nine validators to sit on the same compromised side of a door, and then quietly signed a couple of transfers. This piece walks the timeline and the real reason five of those six relevant nodes fell at once.
Background: Axie Infinity and Ronin
In 2021, Vietnamese studio Sky Mavis ran the breakout play-to-earn game Axie Infinity: pets you bred, battled, and earned the in-game tokens SLP and AXS from. In parts of the Philippines, Vietnam and Indonesia, there was a stretch where families lived partly off Axie earnings.
But Axie’s success collided with an old problem: Ethereum gas was expensive. To keep the game accessible, Sky Mavis built its own sidechain, Ronin, to host all in-game transfers and battle logic. Between Ronin and Ethereum mainnet sat a cross-chain bridge: players locked ETH and USDC on Ethereum, the bridge minted matching tokens on Ronin. The bridge’s security depended entirely on a set of validator nodes that multi-signed every cross-chain withdrawal. It rhymes with the difference between CEX and DEX: you weren’t trusting code so much as a specific set of people and keys.
Nine validators, five signatures gets you through
The Ronin bridge’s setup was: nine validator nodes total, and five signatures were enough to authorize a withdrawal. By industry standards that threshold isn’t outrageous. But here’s the detail: Sky Mavis itself directly operated four of those nine nodes. So an attacker only had to compromise one additional outside node to reach the quorum.
It got worse. One of the remaining five validators was Axie DAO. Back in November 2021, during a player traffic surge, Axie DAO had temporarily authorized Sky Mavis to sign on its behalf. That delegation was never revoked. So on paper you needed five independent nodes; in practice, capturing Sky Mavis’s four sets of keys gave you five votes.

How the attack happened: spear-phished keys
Step one, social engineering. A senior engineer at Sky Mavis was cultivated for weeks over LinkedIn by what appeared to be a recruiter from a non-existent company. Multiple rounds of “interviews,” and finally a “job offer” PDF.
Step two, a poisoned document. The engineer opened the PDF on a corporate machine. The document executed malicious code and gave the attackers an initial foothold inside Sky Mavis’s internal network.
Step three, lateral movement to validator nodes. From that foothold, the attackers obtained access to the four Ronin validator nodes operated internally by Sky Mavis.
Step four, exploiting the dangling Axie DAO delegation. The same captured keys could also sign on its behalf, giving the attackers a fifth valid signature.
Step five, two large withdrawals totalling roughly 625 million dollars at the time. No zero-day. No contract logic bug. Structurally this belongs to the same family as approval phishing and fake support scams, only the target was an engineer with the right keys.
Why nobody noticed for six days
The Ronin bridge at the time didn’t have automated reconciliation alerts. It relied on user behavior: somebody complaining that a withdrawal wouldn’t go through. The hack happened on March 23. The attackers didn’t immediately cash out the USDC, and the contract balance changes got mixed in with normal activity. On March 29, a player tried to withdraw 5,000 ETH from the bridge, found no funds, and pinged Sky Mavis. The team realized the bridge had been drained for six days.
That window gave the attackers nearly a full week of unhurried laundering through dozens of wallets, Tornado Cash, and cross-chain hops. Chainalysis and the FBI later attributed the attack to Lazarus Group, the North Korean state-affiliated outfit behind a long string of crypto thefts.

The aftermath
Sky Mavis paused the bridge and Katana DEX, raised emergency capital (a16z led a 150 million dollar round), and committed to fully reimbursing users. The validator set expanded from 9 to 21+ with independent operators (Binance, Animoca, Ubisoft). The lingering Axie DAO delegation was finally revoked. Real-time monitoring went in.
For ordinary users the takeaway is simple: once funds enter any bridge, your safety is the safety of that bridge’s validator set. It lines up with the same question you’d ask when choosing an exchange.
Why five of six relevant nodes all fell
Strip the story to its core and the fatal flaw wasn’t “an engineer got phished.” The fatal flaw was the structure of the validator set.
First, the nodes were concentrated in one company. Four of nine were operated by Sky Mavis. A single breach surrendered 44% of the network’s signing power. Same internal network, same engineering team, same credential store.
Second, a temporary authorization quietly turned into a permanent backdoor. The Axie DAO delegation should have expired the moment congestion eased; instead it sat forgotten in a config file. That one dangling delegation turned a 4 + 1 problem into a 5 problem.
Third, there was no routine rotation or auditing of signing authority. A validator set isn’t a one-time configuration; it’s an asset that needs scheduled audits and least-privilege defaults.
Together these collapsed “5 of 9 multisig” into something close to “1 of 1.” It’s the same line every cross-chain bridge security report has repeated since: the strength of a multisig isn’t the threshold number; it’s whether the keys behind it are actually independent. When a project advertises “multi-signature custody,” the real question isn’t how many signatures, but whether those signatures live on the same network, with the same operators, behind the same forgotten delegations.
This article is for education only and is not financial advice. Crypto is volatile and risky — only ever risk what you can afford to lose.