The white paper assumes that you don't have a malicious device. If someone can modify the code running your device, they don't need to double spend, they can just directly steal your cryptocurrency.
Yes, there are other ways of stealing besides double spends. We're in agreement there.
And I think we can both further agree that someone that isn't us modifies the code of the Bitcoin node we run on our devices. Unless you're still running Bitcoin 0.1, but then you're actually not on the "Bitcoin" network as we know it today. There are pressures outside of your control that make it so that you have to update your node, and you have to run new code you probably haven't reviewed or approved yourself.
The only way to make Bitcoin work as defined by Satoshi is to have the code be 100% perfect and then not touch it, and that ship sailed a long time ago.
What is this supposed to be a detriment relative to? If someone compromises your device then they can also transfer all of the money out of your bank account, use it to buy gold bullion and be on a beach in a non-extradition country before you get your next bank statement.
"Your bank will refund you for fraudulent charges" doesn't prevent fraud, it puts the cost of fraud insurance into the cost of banking. Which may be significantly less efficient when the user or anyone other than the bank is the lowest cost avoider for preventing fraud.
They also won't refund you for fraudulent charges unless you report them quickly enough.
You've drifted off-course. We were talking about whether or not double spends are possible in Bitcoin and now you're arguing about how banks deal with fraud refunds. peace!
It prevents double spends under a set of assumptions. One of the assumptions it makes is that the Bitcoin code on your device faithfully implements the system in the paper. It doesn't claim to be the thing that causes that assumption to be valid; you have to ensure that using separate means.
Therefore, Bitcoin code has to have been always free of bugs (malicious or accidental). That has not been the case and will not be the case. The assumption doesn't hold, and double spends are therefore possible.
