Matthew Garrett

It'd be pretty easy to avoid the obvious cases - block anything that calls XSelectInput() or XTestFakeKeyEvent(), but people aren't obliged to call the library functions to do this. You could write some pretty obfuscated code that would end up having the same result, and that would filter through any kind of checks that could reasonably be implemented. It's a hard problem - the only way to really handle it is to design the technology such that these attacks aren't possible in the first place.