What is your idea about fractional reserve practice at bitcoin exchanges?

[itsAj]

Fractional reserve exchanges are a terrible idea.  I think all of them should be required to have audits showing they have 100% of everybody's coins.  Leveraged trading is okay if everything is transparent like Bitfinex, but not just making BTC out of thin air.

The only way that bitfinex can lend bitcoin is by borrowing bitcoin from it's other users (it would be too expensive to lend out their own bitcoin).

Bitfinex is essentially working on a two tier system. One - 100% reserves for people who simply have their bitcoin on deposit there. Two - fractional reserve for people who have invested in swaps, these people receive some amount of interest in return for being on the fractional side.

[Brewins]

Banks use fractional reserves because they invest and borrow money, so they gain money with the customer's money.


But how an exchange can earn money with customer's funds? They don't borrow money, also they don't have the same level of protection from the State as banks.

Can't see it not ending as Gox.

[johnyj]

Banks use fractional reserves because they invest and borrow money, so they gain money with the customer's money.

But how an exchange can earn money with customer's funds? They don't borrow money, also they don't have the same level of protection from the State as banks.

Can't see it not ending as Gox.

They don't need customer funds to earn money, they just need to create their own virtual funds

A simple example: I'm an exchange and I create an account with 1 million bitcoins, then I dump those coins to drive the price down from $500 range to $50 range, then I buy back all of those coins and drive the price back to $500 range. This account still have 1 million bitcoins but also have 1 million x $100 = $100 million dollars (Suppose the average profit is $100 during the sell-off). This is a very profitable business, I don't see the reason why exchanges don't want to do it?

From exchange's perspective, the only risk is: When they sell the coins, there might be large buyers absorbed all their coins thus force them to buy back at a higher price. But exchanges see all the clients' data, as soon as they see the large amount of customer deposit is arriving, they could always abort the operation before those fiat money are credited to the customer's account. They could even delay the deposit process to make more room for themselves

[Nagle]

They don't need customer funds to earn money, they just need to create their own virtual funds

A simple example: I'm an exchange and I create an account with 1 million bitcoins, then I dump those coins to drive the price down from $500 range to $50 range, then I buy back all of those coins and drive the price back to $500 range. This account still have 1 million bitcoins but also have 1 million x $100 = $100 million dollars (Suppose the average profit is $100 during the sell-off). This is a very profitable business, I don't see the reason why exchanges don't want to do it?

From exchange's perspective, the only risk is: When they sell the coins, there might be large buyers absorbed all their coins thus force them to buy back at a higher price. But exchanges see all the clients' data, as soon as they see the large amount of customer deposit is arriving, they could always abort the operation before those fiat money are credited to the customer's account. They could even delay the deposit process to make more room for themselves

This is known as "speculating with customer funds". It's a form of theft. The US SEC routinely has people prosecuted for this.

[justusranvier]

They should have third party software performing the Merkle construction once a day lets say.  A second "third party program" could verify individuals account holdings (crypto only mind you) by checking hash values.  This program would be encrypted and stored on exchange servers by a third party.  Then individuals could be assigned personal private keys to decrypt and run the verification software as they please.  The private keys could be stored locally on customer devices.

Encrypting and MACing the verification software would ensure the exchange themselves could not alter this software.  As long as enough users verified their holdings periodically they might be able to prove the exchange was not holding less coins than that declared by the daily merkle tree.

This is not an audit of course, more like an inventory check.  I am I missing something?

We can do much, much better than that.

Exchanges can work together form m-of-n multisig pool in which to store customer deposits, then audit each other continually in real time, provided that they can provide cryptographically secure proof of liabilities.

This arrangement is called a voting pool.

Proof of solvency alone is pretty much pointless, since a fraudulent exchange can prove solvency right up until they decide to take the bitcoins and run.

The voting pool arrangement means that no exchange can unilaterally refuse to honor withdrawal requests - they have to convince a majority of the exchanges simultaneously.

So once they are deployed we should see a substantial reduction in the amount of exchange fraud, at least on the Bitcoin side (there's no such thing as multisig for USD).

[leannemckim46]

They don't need customer funds to earn money, they just need to create their own virtual funds

A simple example: I'm an exchange and I create an account with 1 million bitcoins, then I dump those coins to drive the price down from $500 range to $50 range, then I buy back all of those coins and drive the price back to $500 range. This account still have 1 million bitcoins but also have 1 million x $100 = $100 million dollars (Suppose the average profit is $100 during the sell-off). This is a very profitable business, I don't see the reason why exchanges don't want to do it?

From exchange's perspective, the only risk is: When they sell the coins, there might be large buyers absorbed all their coins thus force them to buy back at a higher price. But exchanges see all the clients' data, as soon as they see the large amount of customer deposit is arriving, they could always abort the operation before those fiat money are credited to the customer's account. They could even delay the deposit process to make more room for themselves

This is known as "speculating with customer funds". It's a form of theft. The US SEC routinely has people prosecuted for this.

This is one thing that the NY regulations are going to try to prevent via audits of balances in customer accounts and audits of bitcoin balances. It would increase costs a little bit however it would result in more security for the customer

[snappa4ever]

They should have third party software performing the Merkle construction once a day lets say.  A second "third party program" could verify individuals account holdings (crypto only mind you) by checking hash values.  This program would be encrypted and stored on exchange servers by a third party.  Then individuals could be assigned personal private keys to decrypt and run the verification software as they please.  The private keys could be stored locally on customer devices.

Encrypting and MACing the verification software would ensure the exchange themselves could not alter this software.  As long as enough users verified their holdings periodically they might be able to prove the exchange was not holding less coins than that declared by the daily merkle tree.

This is not an audit of course, more like an inventory check.  I am I missing something?

We can do much, much better than that.

Exchanges can work together form m-of-n multisig pool in which to store customer deposits, then audit each other continually in real time, provided that they can provide cryptographically secure proof of liabilities.

This arrangement is called a voting pool.

Proof of solvency alone is pretty much pointless, since a fraudulent exchange can prove solvency right up until they decide to take the bitcoins and run.

The voting pool arrangement means that no exchange can unilaterally refuse to honor withdrawal requests - they have to convince a majority of the exchanges simultaneously.

So once they are deployed we should see a substantial reduction in the amount of exchange fraud, at least on the Bitcoin side (there's no such thing as multisig for USD).

I think this would essentially make it impossible to create new customer deposit addresses more then once per day (or would be limited as to how many they would be able to create). This would potentially create negative customer experiences across all bitcoin exchanges.

I think it would also create a false sense of security for keeping bitcoin at exchanges as multiple exchanges could potentially work together to be able to run away with their customer deposits.

[justusranvier]

I think this would essentially make it impossible to create new customer deposit addresses more then once per day (or would be limited as to how many they would be able to create).

No, it wouldn't. Why would you think it would?

I think it would also create a false sense of security for keeping bitcoin at exchanges as multiple exchanges could potentially work together to be able to run away with their customer deposits.

Right now, exchange operators need to cooperate with nobody in order to run off with customer deposits. This would be an improvement.

[spazzdla]

It will lead to the bust of many more exchanges and the theft of many BTC.
Our job is when it happens to teach people this is the correct way of how things operate and the gov should NOT be there to bail out bad players that have been raping you.