FTX held less than $1B in liquid assets against $9B in liabilities

First, it is absolutely better to hold your own keys.

That said, I don't think it's fair to say "nothing" in a situation where holding the keys means having control of the coins.

FTX's problem wasn't that they had liabilities. It was that they didn't have cryptographic custody of the assets they claimed to have. They don't have the ability to make a case in court that the funds they held belong to customers and, thus, should not be considered assets that could be given to their creditors, because they don't have the assets at all.

Kraken not only can prove that they have all the customer funds in the audited currencies, but also which account they belong to.

If individual accountholders can prove Kraken holds not just "all customer coins" but "my specific account's coins," Kraken can also make that case in court.

Whether it holds up is obviously not guaranteed, but that's not "nothing" compared to FTX's situation.

I'm not very worried about the cryptographic security of the Bitcoin blockchain.

I am worried that in 5 years we will learn that some hardware wallets used side channels to transfer bits of your private key out to make it easier to guess for someone who worked at the manufacturer.

You can use an hardware wallet, protected by a HSM, on a computer which is itself airgapped. I've written a decoder which decode Ethereum transactions and verify that at least it's what the hardware wallet says it is signing that it is actually signing (amount / fees / destination address).

So you take your hardware wallet, you connect it to a fully airgapped computer (one without any WiFi capability and without any ethernet whatsoever) (btw let's please not get carried away with exfiltration through "fan rotating speed" or the like and hence that not being a really "airgapped computer" and the very concept of "airgapped" being non-existent).

You then sign a tx on your hardware wallet, which generates a text file. You copy that text file to a USB stick. You check that USB stick from another airgapped computer running the tx decoder software. You can see what's signed.

If it's what you wanted, you broadcast the transaction.

This is reasonably secure.

I'm talking about security for people protecting millions in assets, not $1 K, not $1 billion.

There are still several issues. For example the Ledger hardware wallets, often regarded as the be-all / end-all of hardware wallets require constant updating, needing the wallet to be connected to a computer connected to the Internet to download updates.

You can update the firmware before entering your key (for example on a new wallet), but you cannot install the "Nano apps" before entering your keys. Which is an issue in itself.

Data exfiltration through non-deterministic signatures is another very serious issue.

I haven't looked into using the same seed from different hardware wallet vendors and verifying that you get the transaction signature: if that can be done, I'm all ears.

The Ledger CTO and Ledger overall will constantly dodge questions on these issues.

The answer is basically: "Trust us, we won't exfiltrate your seed through non-deterministic transactions" and "Trust us, we won't exfiltrate your seed during apps or firmware updates".

Firmware updates which aren't even signed with a signature people can verify: Ledger can decide to serve, if they want to, a backdoored firmware leaking seeds through non-deterministic to one person in one thousand if they want to.

And they pretend there's nothing to worry about.

What Ledger should do is let people download firmwares and Nano apps offline, put them on USB keys, and then update their hardware wallets from an airgapped computer.

This would at least allow people to crosscheck their firmware and Nano app hashes.

It still wouldn't solve all the issues.

It's very hard to have something you can really trust and the hardware wallets vendors are really trying very hard to make sure you cannot verify what they're doing.

The two sentences

> Instead of selling and crashing the price

and

> these are held onto and used as a collateral to borrow against.

Are not compatible. If selling the assert crashes its price, it is a terrible collateral.

I would like to see how lenders justified in writing accepting large quantities of FTT as collateral.

From the Kraken terms of service:

> None of the Digital Assets in your Kraken Account are the property of Payward. Payward does not represent or treat assets in your Kraken Account as belonging to Payward. However, a court may disagree with Payward’s treatment of your assets and subject them to claims of Payward’s creditors.

This will be tested in court. Kraken's liabilities matter.