Comment by quotemstr
People demonize attestation. They should keep in mind that far from enslaving users, attestation actually enables some interesting, user-beneficial software shapes that wouldn't be possible otherwise. Hear me out.
Imagine you're using a program hosted on some cloud service S. You send packets over the network; gears churn; you get some results back. What are the problems with such a service? You have no idea what S is doing with your data. You incur latency, transmission time, and complexity costs using S remotely. You pay, one way or another, for the infrastructure running S. You can't use S offline.
Now imagine instead of S running on somebody else's computer over a network, you run S on your computer instead. Now, you can interact with S with zero latency, don't have to pay for S's infrastructure, and you can supervise S's interaction with the outside world.
But why would the author of S agree to let you run it? S might contain secrets. S might enforce business rules S's author is afraid you'll break. Ordinarily, S's authors wouldn't consider shipping you S instead of S's outputs.
However --- if S's author could run S on your computer in such a way that he could prove you haven't tampered with S or haven't observed its secrets, he can let you run S on your computer without giving up control over S. Attestation, secure enclaves, and other technologies create ways to distribute software that otherwise wouldn't exist. How many things are in the cloud solely to enforce access control? What if they didn't have to be?
Sure, in this deployment model, just like in the cloud world, you wouldn't be able to run a custom S: but so what? You don't get to run your custom S either way, and this way, relative to cloud deployment, you get better performance and even a little bit more control.
Also, the same thing works in reverse. You get to run your code remotely in a such a way that you can trust its remote execution just as much as you can trust that code executing on your own machine. There are tons of applications for this capability that we're not even imagining because, since the dawn of time, we've equated locality with trust and can now, in principle, decouple the two.
Yes, bad actors can use attestation technology to do all sorts of user-hostile things. You can wield any sufficiently useful tool in a harmful way: it's the utility itself that creates the potential for harm. This potential shouldn't prevent our inventing new kinds of tool.
> People demonize attestation. They should keep in mind that far from enslaving users, attestation actually enables some interesting, user-beneficial software shapes that wouldn't be possible otherwise. Hear me out.
But it won't be used like that. It will be used to take user freedoms out.
> But why would the author of S agree to let you run it? S might contain secrets. S might enforce business rules S's author is afraid you'll break. Ordinarily, S's authors wouldn't consider shipping you S instead of S's outputs.
That use case you're describing is already there and is currently being done with DRM, either in browser or in app itself.
You are right in the "it will make easier for app user to do it", and in theory it is still better option in video games than kernel anti-cheat. But it is still limiting user freedoms.
> Yes, bad actors can use attestation technology to do all sorts of user-hostile things. You can wield any sufficiently useful tool in a harmful way: it's the utility itself that creates the potential for harm. This potential shouldn't prevent our inventing new kinds of tool.
Majority of the uses will be user-hostile things. Because those are only cases where someone will decide to fund it.