I'm a little confused. I understand that this is a client-side attack on the games, and as such, it's pretty obvious that it is possible to modify the client machine, which is totally in the cheater's control, to do all sorts of things to manipulate the game and prevent the anti-cheat code operating. After all, with this level of access, you could do anything, including (for open systems) running their own kernel. There ain't no way that a user-land anti-cheat system is going to prevent that.
But looking at the paper, at one point they are talking about Direct3D and DLLs, which is mainly Windows terminology, and then they dive of to describe a Linux attack. Maybe they are trying to show that problem spans OSs, although I did not see a reference to that.
There is another way of preventing this type of attack, although it brings back something that I was hoping was dead.
If the hardware/OS/games are created using the generally hated (at least here) concepts proposed by Trusted Computing Group (previously known as the TCPA and the previous Microsoft Palladium project), it would be possible to implement a hardware and software stack that would prevent client side privileged access to the system unless it was signed by a recognised key. This would at a stroke prevent almost all of this type of client side attack, but at the same time would wrest almost total control of a machine from it's owner, making it a data appliance rather than a PC.
Because the detail in the paper is so scant, it looks to me like it is a scaremongering piece to bring security back into focus, to try to allow vendors of software to take more control of the PC away from it's owners.
Where's the tin-foil hat. I think I need it now.