I’m not sure about Windows specifically, I just know you can load your own keys onto the mobo. In general, a cryptographic signature is just metadata tacked onto a file, so presumably yes, you could sign the kernel yourself and load your key so Secure Boot works.
The way Linux distros generally work (e.g. Debian) is to use a shim binary and chain load into their own kernel binary. An exerpt:
Starting with Debian version 10 (“Buster”), Debian supports UEFI Secure Boot by employing a small UEFI loader called shim which is signed by Microsoft and embeds Debian’s signing keys. This allows Debian to sign its own binaries without requiring further signatures from Microsoft.
So even if signing the Windows kernel doesn’t work (I don’t see why it wouldn’t), you could use a loader shim like Debian does to not require loading your own keys.
To be fair, I haven’t read the details of Secure Boot specifically to know how it’s done, I’m just going based on my understanding of PGP (about how signing works), early kernel boot, and high level details about Secure Boot. I’m sure someone sophisticated enough to design kernel-level game cheats could figure out how to make Secure Boot happy without a ton of effort from users.
Secure Boot isn’t designed to prevent users from doing things, it merely prevents malicious code from being loaded at boot (i.e. code that doesn’t have access to the keys loaded onto the Secure Boot module).
That’s quite an important omission because we’re talking about Windows. Windows won’t run kernel or driver that’s not using expected certificates, what would be the point otherwise?
Again, I don’t know the specifics about Windows, so I can’t say exactly what a cheater could or could not do. I do know that kernel chaining does work w/ Windows, otherwise the GRUB bootloader would be DOA.
Whatever Windows does is a completely separate thing from Secure Boot, since Secure Boot only impacts early boot (i.e. the handoff from UEFI to the kernel). So getting into what Windows does and does not allow isn’t particularly relevant to the discussion about Secure Boot.
But that’s completely separate from my point. Regardless of the solution they pick for that particular problem, users can still add their own keys to Secure Boot and do whatever they want.
I’m not sure about Windows specifically, I just know you can load your own keys onto the mobo. In general, a cryptographic signature is just metadata tacked onto a file, so presumably yes, you could sign the kernel yourself and load your key so Secure Boot works.
The way Linux distros generally work (e.g. Debian) is to use a shim binary and chain load into their own kernel binary. An exerpt:
So even if signing the Windows kernel doesn’t work (I don’t see why it wouldn’t), you could use a loader shim like Debian does to not require loading your own keys.
To be fair, I haven’t read the details of Secure Boot specifically to know how it’s done, I’m just going based on my understanding of PGP (about how signing works), early kernel boot, and high level details about Secure Boot. I’m sure someone sophisticated enough to design kernel-level game cheats could figure out how to make Secure Boot happy without a ton of effort from users.
Secure Boot isn’t designed to prevent users from doing things, it merely prevents malicious code from being loaded at boot (i.e. code that doesn’t have access to the keys loaded onto the Secure Boot module).
That’s quite an important omission because we’re talking about Windows. Windows won’t run kernel or driver that’s not using expected certificates, what would be the point otherwise?
Again, I don’t know the specifics about Windows, so I can’t say exactly what a cheater could or could not do. I do know that kernel chaining does work w/ Windows, otherwise the GRUB bootloader would be DOA.
Whatever Windows does is a completely separate thing from Secure Boot, since Secure Boot only impacts early boot (i.e. the handoff from UEFI to the kernel). So getting into what Windows does and does not allow isn’t particularly relevant to the discussion about Secure Boot.
You mean this certificate? The one which will expire next year and leave many old machines with Secure Boot enabled, unbootable?
Unbootable w/o changes, yes, assuming hardware vendors actually respect the expiration date.
But that’s completely separate from my point. Regardless of the solution they pick for that particular problem, users can still add their own keys to Secure Boot and do whatever they want.