Japanese exchange DMM Bitcoin recently lost 4503 BTC, worth over $300m.
So what happened? Did North Korea hack their mainframe? Perhaps a team of elite thieves executed a series of elaborate heists to exfiltrate multisig keys from DMM's vaults?
Yep, the stolen funds were sent to an address that matches the first 5 and last 2 characters of the address DMM routinely uses to handle withdrawals from this wallet.
Classic address poisoning* attack, right?
Well, maybe not.
Remember the crazy fee? DMM has never overpaid like that before from this wallet.
And what are the odds that a poisoned withdrawal just so happened to be sweeping an unprecedented 90% of the total balance in a single transaction?
This was a 2-of-3 multisig holding hundreds of millions of dollars of Bitcoin.
Unless DMM is completely insane, each of those keys will be controlled by a separate person.
So imagine you gain access to just one of those keys, but want to steal all of the money.
What do you do?
Perhaps you craft a consolidation to a confusingly similar address, and send it off to another keyholder for approval.
And then hope they don't compare the actual destination with their whitelist of safe addresses too carefully before signing the transaction.
Which leaves some obvious questions:
How did the attacker get access to even one key?
And how did they know who and how to contact to secure that crucial second signature?
I suspect we'll never know...
h/t to my eagle-eyed anonymous source for spotting the address similarities!
*("address poisoning" is when an attacker sends funds from an address they control, which has been "mined" to look confusing similar to one of the victim's own, in the hope that they'll accidentally copy and paste the wrong address from their transaction history)
• • •
Missing some Tweet in this thread? You can try to
force a refresh
seems to be driven by lots of transactions like this, committing to a large batch of inscriptions, but also creating a bunch of seemingly pointless 420-sat outputs. mempool.space/tx/010b7b61208…
the inscriptions that have been revealed so far are recursive references to some 3D orb model.
so I assume this is another mint event for one of these vaporware metaverse projects.
So while this might sound like good news for @83_5BTC, there's a serious risk that AntPool ends up returning the overpaid fees to the hacker, not the victim.
(especially if they just send it right back to the allegedly compromised address 😱)
3/ The most likely explanation is that the wallet was generated from bad entropy - perhaps a brainwallet, or too few dice rolls/coin flips, or an insecure PRNG.
Let this be a reminder not to take shortcuts with your entropy, and ideally to use multisig for very large sums.
I had initially discounted that possibility, but after receiving a tip-off I took another look.
The overpaid fee came from a hot wallet reusing the address bc1qr3...zpw3, which started operating in June of this year.
The on-chain activity is consistent with automated processing of fiat-denominated withdrawals, and also closely matches the behavior of a now inactive wallet bc1qhs...kx4n, which is labelled as PayPal on .
a substack post going around at the moment claims that a single entity owns 64% of all inscriptions created since early March, paying an eye-watering 1056 BTC for the privilege
I've seen a lot of takes already suggesting this sounds like market manipulation, money laundering, or a well-funded attack on Bitcoin by wealthy adversaries.
but the truth is much less exciting.
inscriptions are created with a two-phase commit/reveal process.
first, a taproot output is created which commits to the inscription data and a public key.