Researcher Florent confirmed with The Block on May 31 that he helped recover about 1,003 ETH (about $2 million) that had been trapped in the 2016 HongCoin ICO contract for nine years. As of the time of reporting, two investors have already claimed a total of 96.5 ETH and have voluntarily paid Florent a white-hat bounty.
Confirmed technical reasons and solutions for the integer overflow vulnerability
Florent confirmed to The Block that the HongCoin contract was deployed using an older version of Solidity and lacked protections against integer overflow (later fixed through the SafeMath library). The refund function refused to process token balances held by anyone whose holdings exceeded the global counter. Over the years, partial refunds reduced the counter to 356, causing the refund cap to be only 3.56 ETH (about $7,000), while the balances of most holders were far beyond that value.
Florent confirmed that the solution was to use the HongCoin team’s administrative functionality—originally used to mint bounty tokens—which, when called with specific input values, resets a holder’s balance to 1 due to the lack of overflow protection. The refund check then passes immediately, releasing the locked ETH.
Not a unilateral attack: the HongCoin team signed the unlock transactions themselves
Florent confirmed this was not a unilateral hacker action. He first sent an email to the HongCoin team, verified the operation sequence number on a Foundry mainnet fork, and then the HongCoin team signed the unlock transactions themselves. From sending the first email to completing the entire process took about a week. There were 41 holders who needed their balances reset, corresponding to about 1,000 truly frozen ETH, while the remaining 7 holders could claim refunds directly; the team signed 41 transactions in total.
Florent confirmed the reason the hacker previously ignored the contract was that there is “no ownership vulnerability in the contract that would allow the hacker to steal funds—any attack’s only outcome is that the ETH is returned to the original investors,” meaning there was no profit opportunity for the hacker.
Confirmed methodology: the scanner and known limitations of Claude Code
Florent confirmed that he set up a self-hosted Ethereum node and ran a scanner to flag contracts holding more than 100 ETH, then filtered them one by one. He confirmed he used Claude Code to accelerate contract sorting and clustering, but he said AI has limitations when analyzing specific smart-contract vulnerabilities: “AI is often affected by the fact that the contract hasn’t been broken before, so it often defaults to ‘this can’t be broken,’ and that’s often wrong.”
Florent confirmed on May 24 that he previously also recovered 19.329 ETH (about $40,590) from two older contracts, including a failed ICO contract from 2018 and user funds after Liquality closed its app in 2024.
FAQ
How did the HongCoin contract integer overflow vulnerability cause funds to be locked for 9 years?
Based on Florent’s confirmation, the refund function requires the holder’s token balance not to exceed the global counter. Over years, partial refunds reduced the counter to 356, lowering the maximum refund amount to 3.56 ETH, while the balances of most holders were far above that value. A lack of integer overflow protection in the older Solidity version was the root technical reason; the SafeMath library later fixed the issue.
Why didn’t the earlier hacker attack this contract holding a large amount of ETH?
According to Florent’s confirmation, there is no ownership vulnerability in the contract that would let an attacker steal funds; any attack would only result in ETH being returned to the original investors, leaving no profit opportunity for the hacker, so there was no attack incentive.
How was Claude Code used in this white-hat action, and what limitations does AI have?
Florent confirmed that Claude Code was used to speed up contract sorting and clustering, but he also confirmed that AI has limitations when analyzing smart-contract vulnerabilities. It is often influenced by the fact that the contract had not been broken before and therefore defaults to being unable to break it; he noted that this judgment is often wrong.
