Hacker News Reader

Comment by amelius | Hacker News Reader

That's why I inject noise into the web on a regular basis. Just do some random searches, click some random links.

Let's say there are 10,000 subjects and you are interested in 10. You click 100 links for the subjects you are interested in, and also 1,000 links picked at random. Now random subjects get somewhere between 0 and 2 clicks, way less than your real interests.

You could maybe pick 20/30/40 random subjects to fake a prolonged interest in, and really focus on those, maybe swapping some out over time. In any case, I think one has to pump enough fake volume through to make fake interest look like real interest. But then the opposite problem comes up - excessive requests for one topic might signal fakeness. Maybe you could have fake traffic be produced proportionally to real traffic?

Advertisers would still win, though. Instead of showing you ads for 10,000 subjects of which one interests you, they are now showing you ads for 2 subjects of which one interests you. Much better targeting.

And if they want to link your profiles from elsewhere, two interests of which one is fake is still pretty powerful.

From the presentation:

> Can I hide in my data by generating noise? (e.g. via random page visits)

> Usually not

what about adnauseum ? my browser appears to click every as I am exposed to, but I don't see them.

better or worse?

I'd imagine that is much worse. It's not random at all, you're basically giving them huge amounts of extra data points. That extension was supposed to used to mess up click rates for ads, not stop you from being tracked.

But in any case it gives plausible deniability.

the problem is that humans are really bad at generating random noise. you'd need an automated solution that makes sure you are generating unformly random data.

Reply on news.ycombinator.com