Bitcoin Gold’s majority attack made its last known attempt on May 19

Bitcoin Gold’s developers later identified May 19, 2018 as the date of the last known attempt in a multiday majority-hashrate campaign targeting cryptocurrency exchanges. The incident mattered because it converted a theoretical weakness of smaller proof-of-work networks into an operational settlement failure: an attacker controlling sufficient computing power could replace confirmed transaction history and try to withdraw exchange assets before an earlier deposit disappeared.

The event-day record was incomplete. Bitcoin Gold had warned exchanges and its community about double-spend attempts by May 18, but the project did not publish its comprehensive public account until May 24. What could be established on May 19 was therefore narrower than what investigators and researchers documented afterward.

What was visible on May 19

A Bitcoin Gold forum warning initially reported on May 18 that the most recent observed attack had occurred roughly six hours earlier. The project subsequently amended that notice after another attempt, identifying May 19 as the date of the last known attack.

The warning described a malicious miner privately building an alternative chain while depositing BTG at an exchange on the public chain. After the exchange credited the deposit, the attacker could trade the BTG and withdraw another asset. Publishing the privately mined chain could then erase the original BTG deposit from the accepted history while leaving the withdrawn asset beyond the network’s reach.

That mechanism did not allow the attacker to spend coins belonging to arbitrary users or create valid signatures for other wallets. It exploited the exchange’s willingness to treat a deposit as settled before the cost of replacing its supporting blocks became prohibitive.

Why exchanges carried the risk

Automated exchanges were attractive targets because they combined large deposit capacity, liquid order books and withdrawals into assets operating on other blockchains. Once a withdrawal left in bitcoin or another cryptocurrency, reorganizing Bitcoin Gold could not reverse that separate network’s transaction.

Bitcoin Gold advised exchanges to increase the confirmations required before crediting BTG deposits. A longer confirmation requirement forces an attacker to construct a deeper alternative chain, raising the necessary time and computing expenditure. Closing BTG deposit wallets removed immediate exposure but also interrupted legitimate transfers.

The project’s later account said many exchanges either increased confirmation requirements dramatically or stopped accepting deposits. It also acknowledged an important evidentiary limit: the developers could observe attack attempts and reorganizations but could not determine how many attempts successfully extracted assets or conclusively identify every affected exchange.

The market did not yet have a complete disclosure

CoinMarketCap’s May 19 historical snapshot recorded bitcoin at $8,247.18, unchanged to two decimal places over the publisher’s displayed rolling 24-hour window, while ether was $696.53, up 0.55%. Those aggregated observations provide broad market context only. The surviving page does not specify an exact snapshot time, venue composition or revision history, and neither figure measures Bitcoin Gold’s attack-related impact.

The absence of a broad event-day selloff does not establish that the attack was immaterial. The project’s fuller explanation and outside estimates of the affected value appeared after May 19, meaning much of the market could not have incorporated a complete public account during the measured window.

Why proof-of-work scale mattered

A proof-of-work ledger’s practical resistance to reorganization depends on the computing power already securing it and the cost of obtaining compatible power. Bitcoin Gold used Equihash, an algorithm also used by larger networks. Its developers said the wider pool of compatible mining power exceeded the power normally protecting Bitcoin Gold, leaving the smaller network exposed if enough capacity shifted toward an attack.

This did not show that every proof-of-work chain faced equal risk. Network size, compatible hardware, rental availability, exchange controls and extractable liquidity all affected the economics. It did demonstrate that decentralizing access to mining equipment did not automatically guarantee an immutable transaction history.

Later-record clarification

On May 24, Bitcoin Gold said the attacks had spanned approximately three and a half days and confirmed May 19 as the last known attempt. A 2020 Bank of Canada staff paper subsequently dated the successful double-spending campaign to May 16–19 and cited an approximately $18 million value. That figure is later analytical context, not a loss total that was verified publicly on May 19, and it should not be treated as a complete exchange-by-exchange accounting.