I think this would slow down the process of being able to sell bitcoin on their platform, as well as the ability to to debit customers' losses when they have leveraged losses.
Probably not so much, because they only settled accounts to the blockchain once per day. The exception being, if you bought bitcoins and immediately requested to withdraw them. That already took 15-45 minutes generally. Not sure how much this would really add.
This would prevent a user from being able to place a sale order without first confirming with BitGo. Also, as I previously mentioned, if a customer opens a leveraged position with BTC as collateral, that declines in value, then the customer would presumably not agree to have BTC withdrawn from his account, even though this would be the appropriate thing to do.
The problem with BitGo is that they signed transactions that they should not have signed. I speculate that BitGo was signing transactions they believed to be "internal transfers" between bitfinex accounts, which I suspect are not subject to the limits that bitfinex set, although this is speculation.
They certainly signed transactions they shouldn't have signed. Even if they were internal transfers, that kind of volume (above, say, 20k or 30k) should trigger a circuit breaker. A telephone call, some kind of second step verification...something, rather than simply auto-signing for 120k bitcoins.
I know that BitGo handles internal transfers differently from outright withdrawals, at least in terms of pricing (a review of their website says that internal transfers are free while withdrawals signed by BitGo start at 0.1% with volume discounts available). It would make sense for there to be different limits between internal transfers and withdrawals.
I wasn't referring to trades. I was referring to withdrawals. Finex said everyone on their exchange had a BitGo account. If this is true, Finex, did not use commonsense as I had previously mentioned in terms of withdrawals.
However, you make a great point. If this was done in a way to make it appear as there were trades and simply transfer to settle trades between accounts, that's an entirely different issue to contend with.
I understand that when a BitGo customer will need to create a new "account", the BitGo customer (in this case Bitfinex) will provide BitGo two public keys, and BitGo will provide the customer with one public key, all of which will be combined to create a 2-of-3 multi-sig address. Ideally, when bitfinex provides their two public keys, one of the corresponding private keys will be held in offline cold storage, and one of them will be held in their online/hot server.
I am not entirely sure how bitfinex handles the creation of BitGo wallets/accounts when a new bitfinex account is opened. Ideally, Bitfinex and BitGo have exchanged a large quantity of public keys (tens/hundreds of thousands, or even millions), so when a bitfinex customer opens their account, BitGo can simply "activate" 3 wallets/accounts for this customer. BitGo most likely does not have access to bitfinex's business records, so they will simply be told by bitfinex when a new wallet/account needs to be activated. It is possible however that Bitfinex and BitGo exchange keys at the time the new bitfinex account is created, and the attacker provdied BitGo two public keys, and BitGo provided the attacker one public key, and that the attacker would have access to both of the corresponding private keys. This would allow the attacker to potentially first direct BitGo to sign transactions moving customer BTC to these newly created "wallets" and the attacker subsequently using the two private keys that he is in possession of to create a second transaction spending the BTC to an address that he created far away from the bitfinex servers. So BitGo would have only signed "internal transfer" transactions, and the attacker signed the "withdrawal" transactions.
It appears that the attacker was able to work for several hours without detection, which IMO is a very long time for an unauthorized intruder to have access to a server. Based on the statements of bitfinex (and BitGo), it appears that bitfinex was not initially sure how the attacker was able to access their servers, and based on the fact that users are still not able to access their bitfinex accounts, they still may not know. Based on the above, I suspect that the attacker was likely able to use some kind of 0-day attack.
If you assume that the losses at Gox actually happened a long time before Feb 2014, then this is, by far the largest hack/theft of Bitcoin in history.
Most of this is of course, speculation.
