Doesn't this blocker still require you to load and render the ads so they can apply their vision techniques to them? You won't get any performance win and you'd still see empty popovers or blank space. I don't see how detecting that something is an ad is enough on its own to create an effective ad blocking solution.
The part about defeating known anti-ad blockers seems rather unfair. If this technique did gain popularity I'm sure there would be ways around it (like tricking it into thinking the button used to dismiss a popover is itself an ad).
> Doesn't this blocker still require you to load and render the ads so they can apply their vision techniques to them?
Definitely seems that way from what I understood. Seems like a novel technique, but not something I'd personally flock to use unless my network/element based rules suddenly became ineffective.
Rather than being used via an extension in the client's browser though, this technology looks like it could be incredibly useful for automatically figuring out rule changes and updating e.g. the element hiding rules automatically.
It might be possible that for an ad-blocking system that,
for all visitors to a specific page, vision analysis only has to be applied once on that specific page and share that post-analytic knowledge publicly. Then all future visitors would benefit from that previous analysis knowledge and block the ad without yet another vision analysis on the same specific page.
to scope with pages changing widly, the system can assign 5% of the visitors to each specific page to perform new vision analysis and update the findings.
The part about defeating known anti-ad blockers seems rather unfair. If this technique did gain popularity I'm sure there would be ways around it (like tricking it into thinking the button used to dismiss a popover is itself an ad).