INDEPENDENT CRASH TESTSTEST PROTOCOL PUBLISHED IN FULLRATINGS YOU CANNOT BUYEVERY REPORT DATED & RE-TESTEDPROTOCOLS TESTED ACROSS 12 CHAINSNEW: BRIDGE CRASH SERIES · READ NOW
Request a Test ↗
Home / Reports / Post-Incident
Post-Incident · Public record

Five signatures, one breach.

Five of nine validator keys were enough. The bridge took six days to notice. A public-record review of the Ronin exploit.

7Min read23.03.22Incident5 / 9Signatures

Public-record analysis. Not a BULWARK Crash Report. No star rating is attached to this article.

Incident23 Mar 2022
Detected29 Mar 2022
Stolen173.6k ETH
Also stolen25.5M USDC

What happened.

On 23 March 2022, an attacker used two forged withdrawals on the Ronin bridge, the cross-chain link built for Axie Infinity. The transactions moved 173,600 ETH and 25.5 million USDC to a single wallet. At the ETH prices used in public reporting, the loss was roughly $540 million to $625 million.9

Ronin required five of nine validator signatures. The attacker obtained five: four from Sky Mavis validators and one from an Axie DAO validator. The bridge checked the threshold and processed the transfers. There was no secondary control between a valid signature set and a payout.312

“A permission we stopped using is a permission nobody has”—except it isn’t.

The fifth signature came through an Axie DAO allowlist that had been used for a temporary delegation in late 2021. It was never revoked. The attacker could still ask the DAO validator to sign.332

Six days of silence.

The theft was not detected internally. It surfaced on 29 March when a user reported that a 5,000 ETH withdrawal would not go through. Sky Mavis then disclosed the breach and halted the bridge and Katana DEX.936

Sky Mavis’s own post-mortem acknowledged the missing control: there was no monitoring or alerting for large bridge outflows. No automated circuit breaker, and no anomaly check on transaction volume. A user complaint was the detection system.32

That is the part a rating should capture. A protocol can pass a contract review and still fail the monitoring test. The loss did not come from a broken signature check. It came from a valid signature set with no independent reason to stop and look.

What the Pole Test would ask.

BULWARK’s Pole Test assumes one privileged key is lost or misused. Ronin shows the more useful question: what happens when the threshold itself is reached through a combination of keys that should no longer be able to sign together?

  • Are all signer permissions current, and is every delegation revoked on-chain with an expiry?
  • Can one compromised operator reach the threshold, or does the set require independent organisations?
  • Does a large withdrawal pause for human review before settlement?
  • Does the bridge alert on outflow size, velocity, and destination before the funds leave?

A protocol that can answer those four questions with evidence is in a different risk class from one that can only show a multisig address. The difference is not cosmetic. It is the difference between a threshold and a control.

Attribution, recovery, and the redesign.

On 14 April 2022, the FBI attributed the theft to the Lazarus Group and APT38, North Korean state-sponsored actors. OFAC sanctioned the attacker’s Ethereum address the same day.16

Sky Mavis raised $150 million, led by Binance, to reimburse users. In September 2022, Chainalysis and U.S. law enforcement recovered about $30 million, roughly 10% of the stolen funds at adjusted prices.24

The bridge reopened on 28 June 2022 with a redesigned withdrawal process, including tiered reviews and a human check for the largest withdrawals. The design change is the honest part of the story: the team did not just add another signer. It added a reason for a large transfer to stop.1733

What this means for a Crash Report.

This article is not a BULWARK test. It is a public-record case study used to sharpen the questions the lab asks. A future Ronin-style report would need the fork block, the contracts reviewed, the signer set at that block, the timelock and monitoring evidence, and a signed verdict from a Test Director. None of that can be inferred from a news article, and none of it is replaced by this piece.

The lesson is narrower and more useful: a threshold is not a monitoring system. A bridge that can move funds in one transaction, with a valid quorum, and no independent pause, is one key-management failure away from a very bad day.

Sources.

  1. FBI — attribution statement, 14 April 2022
  2. Sky Mavis — funding and reimbursement announcement
  3. Halborn — explained: the Ronin hack
  4. Chainalysis — Ronin seizure analysis
  5. OFAC — April 2022 action
  6. Elliptic — Ronin incident analysis
  7. J. Becker — technical analysis of the Ronin attack
  8. Decrypt — bridge reopening
  9. AccessCreep — Ronin 2022 review
  10. Sheng — multisig as a single point of failure
  11. Al Jazeera — Ronin disclosure

BULWARK · Public-record analysis · Not a signed rating · Published 07.11.26