The last 10 years have seen astonishing (and ongoing) progress on recognition problems whereas all analysis of side-channel confinement problems in the same period has only increased our estimates of how costly they are to solve.
1
2
8
Total hand waving here, but I'd guess the cost of "solving" confinement for reasonable user goals around anti-fingerprinting grows exponentially with total system complexity (sw+hw), but recognition + intervention around abuse might grow only logarithmically.
1
1
3
Replying to @hillbrad
Has anyone looked at introducing noise into fingerprints rather than trying to stop it? Eg periodically return wrong screen size, or periodically send bogus UA, etc.

Jun 15, 2019 · 3:02 AM UTC

1
Replying to @bilcorry
I'm aware of suggestions in this vein going back many years. Not sure how practical it is. Classification of real abuse at scale could inform these kinds of mitigations.