How Mt. Gox Imploded
motherboard.vice.com
motherboard.vice.com
I also made a timeline of events, which is much more detailed than the article: http://newslines.org/mt-gox/?order=ASC
It was my understanding that double spending wasn't possible?
This (known) flaw was about being able to change certain parts of a transaction that were unsigned, allowing multiple equally valid transfers to have different ids. Only one of then ends up in the blockchain, however.
You could thus not rely on a transaction id as a reference to see if a payment was made, but some systems did nonetheless.
http://www.theregister.co.uk/2014/04/09/mt_goxs_transaction_...
Karpeles is either a deliberate fraud (he does have a prior history) or possibly a bad programmer whose bugs in code dealing with money caught him with his pants down...
Eitherway he has greatly damaged Bitcoin name with his shenanigans, blaming transaction malleability was one of the straws he tried to clutch at as his house of bad code collapsed, he also tried to blame his staff and few other tricks.
He certainly didn't understand a lot about security. In the original topic on bitcoin talk (https://bitcointalk.org/?topic=444.0) someone pointed out he was sending passwords in the query string. He replied "It is a post over https. It is secure." ignoring the other (obvious) problems with putting passwords in the query string (http://security.stackexchange.com/questions/29598/should-sen...)
Apparently they were using the hash to track transactions, and would re-send bitcoins if the transaction didn't appear to go through. This is obviously a problem if the transaction did go through, but with a different hash.
Basically, two transactions which effectively do the same thing can end up having a different hash. Because of this, one should not rely on looking for the hash in the blockchain to see whether a transaction has "gone through".
This has been a {known issue, flaw} since 2011, but some BTC exchanges (like Mt. Gox) continued to rely on the transaction hash to check whether transactions had been committed to the blockchain. If a transaction hadn't appeared to have gone through, they re-sent bitcoins (under a new transaction). This is a problem if the transaction did indeed go through, but under a different hash, as it would cause the same action (transfer money from A to B) to show up in the blockchain twice.
0. http://www.righto.com/2014/02/bitcoin-transaction-malleabili...
In the case of Mt Gox, poor internal accounting controls made it possible to make repeated withdrawals.
Say a standard transfer is initiated by sending Mt Gox a message, [Mt Gox, send me money.], which causes Mt Gox to check that the funds are available, initiate a legitimate bitcoin transaction and debit the Mt Gox account for that amount. Their system was set up so that if you sent a specially formed message like [Mt Gox, send me MORE money.], it would check that the funds were available, initiate a legitimate bitcoin transaction and not bother debiting the Mt Gox account.
That isn't quite technically accurate (the attack involved getting Mt Gox to respond to a different message than ended up on the consensus blockchain), but it captures where the problem was, Mt Gox wasn't keeping track of funds in a careful way.
While I'm far from authoritative on the matter, under certain circumstances, it appears to be very possible.
Here's one person's explanation of their $10,000 double-spend during a hard fork of the blockchain:
https://bitcointalk.org/index.php?topic=152348.0
A mining pool also manipulated the block chain to run a double-spend attach against a dice betting site:
http://www.reddit.com/r/Bitcoin/comments/1qqmr4/ghashiocexio...
That doesn't sound right to me. I seem to recall that the price on MtGOX was lower than elsewhere.
"This is an excerpt from _Digital Gold: Bitcoin and the Inside Story of the Misfits and Millionaires_, a new book by New York Times reporter Nathaniel Popper. Reprinted courtesy of Harper, an imprint of HarperCollins Publishers. "