One of the threat models is that a fraudster tricks a non-technical user into installing malware, which then manipulates the user interface so that next time the user tries to send money to Bob, it actually goes to Mallory. That's a legitimate concern, and one of the causes why PSD2 mandates that all 2FA devices must have a display that shows the user where they're about to send the money and how much.

> Can you show me examples where locking down an OS has prevented fraud in banking?

This is a non-sensical remark because it's impossible to "prove" a counterfactual. I find stuff like this incredibly fucking annoying - please don't say this.

Look at the last 30 years of computing history?

When online banking was first created it was an absolute chaos zone. Everyone was accessing it from desktop machines riddled with viruses and malware. There are endless stories of being discovering their life savings had been wired to Belarus by some malware running on their machine that had grabbed their banking credentials when they logged in.

https://www.google.com/search?q=site%3Akrebsonsecurity.com+b...

https://krebsonsecurity.com/2017/07/how-a-citadel-trojan-dev...

> U.S. prosecutors say Citadel infected more than 11 million computers worldwide, causing financial losses of at least a half billion dollars.

Half a billion dollars, by a single guy with a single virus!

Different parts of the world came up with different solutions for this. The US made all ACH payments reversible and international wires difficult, but that just meant the receiver paid for fraud instead of the person whose machine was full of viruses. This was an obviously bad set of incentives and hacky panic-based fix. Banks elsewhere in the world settled on providing users with authenticator devices that looked like small calculators into which you could type transaction details after plugging in a smart card. Malware could still steal all your financial data but it couldn't initiate transactions.

Obviously, all this was a hack. What was needed was computers that were secure. Apple and the Android ecosystem eventually delivered this, and the calculator devices were retired in favour of smartphones with remote attestation. This was better in literally every way, for 100% of users. Firstly, it protects financial privacy and not just transaction initiation. Secondly, it's a lot more convenient to use a device that's always with you than a dedicated standalone single-use computer. Thirdly, adding remote attestation made no difference because that's what the calculator devices were doing anyway. Fourthly, even in the case of customers of small American banks that weren't capable enough to manage dedicated hardware rollouts, getting rid of fraud instead of pushing liability around allows for lower prices and fewer headaches.

So remote attestation is a non-negotiable requirement for digital banking of any form. When Microsoft didn't deliver most banks preferred to literally manufacture and sell their customers single-use smartcards that remotely attested by you manually copying numbers back and forth between screens. Or they hid the cost of rampant fraud in the price of other services until such a time that Apple/Google saved them.