Bitcoin bots stole exposed test funds within seconds, Belgian researchers find
Automated programs are waiting on the Bitcoin network for wallets to make a specific, old mistake. When two Belgian researchers made that mistake on purpose, bots tried to take the money within seconds, according to CoinDesk.
Computer scientists Vincent Jacquot and Benoît Donnet of the University of Liège put $3 to $5 each into Bitcoin and Ethereum accounts they had deliberately left vulnerable. On Bitcoin, attacks began "a second or a few seconds after the bait was set up," Jacquot told CoinDesk. At least two rival transactions competed to claim the coins while they sat in the mempool — a waiting area where transactions sit before the network confirms them.
Key points from the research
- Automated attackers targeted deliberately weak Bitcoin wallets within seconds after the researchers exposed a signing error that can reveal private keys.
- A review of six blockchains through June 2024 recovered 3,620 private keys, including 3,278 on Bitcoin, by finding reused signature numbers known as nonces.
- The peak balances of vulnerable addresses totaled about $106 million at November 2024 prices, but the researchers did not establish how much money was actually stolen.
How the wallet mistake works
The flaw is in how Bitcoin wallets sign payments. Every signature uses a random number called a nonce, which is supposed to be used only once. If the same secret key signs two transactions with the same nonce, anyone who sees both can work out the key with basic algebra and take the money.
CoinDesk compares it to reusing the same one-time password for two different bank transactions. Once someone sees both, they may be able to work out the underlying password and take control of the account.
What the test showed
In the experiment, the researchers sent 7,200 satoshis — the smallest unit of bitcoin, worth a few dollars — to each of six newly created addresses from one wallet. About an hour later, two of those addresses paid each other, and both used the same nonce. That alone revealed nothing, because two different keys had signed the transactions.
Then one of the pair signed a third payment with that same nonce. That meant a single key had reused its nonce, and the math needed to unlock it was available for anyone to do. The next transaction out of that address sent its remaining 5,200 satoshis, worth a few dollars, to a stranger.
Why the open network makes this possible
Once a transaction enters the mempool, it is visible to everyone. That lets automated attackers monitor the network around the clock, waiting for predictable mistakes rather than targeting specific victims.
The researchers' finding also shows why wallet security cannot depend on users alone. Developers, wallet providers, and exchanges need safeguards that stop common errors from becoming instantly exploitable by bots, according to the reporting.
What is confirmed and what is not
The researchers' experiment, the recovered key counts, and the $106 million peak-balance figure come from the study and CoinDesk's reporting. How much money was actually stolen from vulnerable addresses was not established by the researchers, so that remains unknown.