The GOAT BitVM3 Bug Bounty ran for six weeks, from August 18th to September 29th. Independent researchers were invited to attack the BitVM3 bridge before it carries mainnet BTC - the node software, the bridge contracts, the transaction library and the proving stack - with a $5,000 reward pool for accepted findings.
Results from the first batch of submissions were published in September. The engineering team has now completed its review of the final batch, which closes out the program. This is the full record.
What the bounty was testing
The BitVM3 bridge is the system that moves BTC between Bitcoin and GOAT Network, and it is the piece of infrastructure everything else depends on. GOAT Network is a rollup: its state is proven back to Bitcoin, and the bridge is what lets value move on the strength of that proof. Building a bridge that holds to Bitcoin's standard is what makes Bitcoin usable as the settlement layer for everything built on top.
A withdrawal from the bridge is not paid out on anyone's say-so. The operator fronting a withdrawal has to prove it followed the rules, the rules it is judged against are committed on Bitcoin rather than in the software they govern, and every exit is held open to challenge for at least a week before it can finalize. A raised dispute locks the exit behind further timelocks and is resolved onchain, with proof generation and dispute resolution powered by Ziren, @ProjectZKM's zkVM. The trust assumption is deliberately small: the bridge stays safe as long as one honest verifier checks the work during the window.
How a withdrawal moves through the BitVM3 bridge. The challenge window and the verifier logic behind it are where the bounty's strongest findings landed.
A design like that assumes software has bugs and builds the recourse in. But the challenge logic, the node that runs it and the contracts that anchor it are themselves software, and a bug there is the one risk the architecture cannot remove on its own. That layer is what this program paid researchers to attack, and it is where the strongest findings landed across both batches: the verifier and challenge paths that make the design work. The full design is in the BitVM3 whitepaper, with the bridge architecture documented here.
The final batch
The closing batch brought 37 classified findings: 2 High, 2 Medium, 11 Low and 22 Informational. Three are receiving awards.
High: a gap in how the bridge validates timelock configuration on its challenge path, affecting the node and the BitVM transaction library
Medium: missing resource limits in the node's peer-to-peer message handling, usable as a denial-of-service vector
Low: a witness validation gap spanning the node and the L2 contracts
For these three we are publishing the affected areas rather than the mechanics. The fixes are being implemented now, and the full details will be visible in the public repositories once they land.
As with the first batch, a number of further submissions were ineligible because they described issues already patched, duplicated other reports, or fell outside the program's scope. Classified findings that did not receive awards were largely informational or below the award threshold.
The program in total
Severity breakdown across both batches, as classified by the GOAT engineering review.
Across the program, 123 classified findings and 16 awards: 5 High, 6 Medium and 5 Low. Researchers covered the whole surface - node and P2P behaviour, verifier and challenge logic, bridge and contract logic, transaction validation, proof and circuit-level analysis, and API behaviour. The five awarded findings from the first batch that arrived as public GitHub issues are documented in the first batch results; the rest came through the private submission form.
With the final review complete, rewards for all 16 awarded findings are being distributed together from the $5,000 pool, as set out at launch. Individual reward amounts are not disclosed.
What the program leaves behind
Every awarded finding becomes a fix in the public repositories, several already closed. Six weeks of adversarial review produced sixteen confirmed issues that will be fixed before the bridge carries mainnet BTC. Work at this layer has to be tested by people trying to break it, and now it has been.
Our thanks to every researcher who took part. The original program scope and resources remain available in the launch announcement.
