Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I think that such strategy with data poisoning will not be effective because based on current data your activity will be marked as anomaly and will not go into any dataset. So this strategy is the same as doing nothing.

But I agree that it's pointless to delete visible data about yourself, because statistics are already stored somewhere else and you don't have power over it.



> based on current data

It might well be useless to start today with your existing accounts. But it's a principle worth knowing about nonetheless. It's certainly possible to poison data from the beginning of your contact with a service - especially for ones less pervasive than Facebook. An account that never sees entirely true data may not be able to build up an accurate profile.

Having said that, providing false information does violate the ToS for most sites. Which probably isn't a federal crime, given US v Lori Drew, but it's certainly not a good idea for an account you can't afford to have suspended.


I wouldn't say its pointless. What if someone gains access to your account through social or technical means? Sure Facebook/NSA/etc have a copy, but at least Customs agents, hackers, and crooks wielding a rubber hose can't access it.


Now there's a business idea.. have an app running on your phone and PC that watches your "normal" facebook activity to establish a baseline, and ever so slightly perverts it over time using the same connection so the IP/location information still lines up. Slowly and methodically so the information doesn't get flagged as bogus. Start with overtly partisan political stuff since that's the easiest to classify.

Call it Social Chaff as a Service (SCaaS).


I have been fantasizing as a combination of data poisoning and intervention when the poisoning is actually real: make groups to swap profiles for a while. Without even going back to the original, but just jumping to the next one. As long as you agree through a different medium and swap credentials with someone you somehow trust, it should work.


why go through all that? who would go through all that? making your own life more difficult in spite of a company? it's not sustainable

note: I know that some people actually would. but philosophically speaking I just don't get it, it's almost like its missing some kind of deeper point.


> I think that such strategy with data poisoning will not be effective because based on current data your activity will be marked as anomaly and will not go into any dataset. So this strategy is the same as doing nothing.

You give them too much credit. This would require some PLM being concerned with such activity enough to add the anomaly detection. Unless they are way proactive over theoretical issues.


Why not build a browser plugin that visits 5 random websites for each site you visit.


https://adnauseam.io/ There you go. Not on the chrome extensions store, but works on chrome (according to their site, I can only vouch for Firefox). I had this question a while ago, and debated building my own, but laziness won out. Edit: I'm an idiot, didn't read that you specified random websites. Adnauseam just clicks on random ads to pollute marketing data.


Why are these people not creating spaceships, exploring the deep of the oceans, or curing cancer?


There is no advertising money in those fields.


Lest you think this is an exaggeration, graduate students and postdocs make peanuts, who would be doing the cancer research, make peanuts.

Starting salary for a postdoc is a bit less than $50k, and this is after 5+ years of grad school, during which you make even less (often $20-30k) One can expect to be a postdoc for 3-5+ years before becoming remotely competitive for an independent (faculty) position, and another 5 years before the job is relatively secure. The competition for these jobs—and the remaining industrial ones—is savage and unpredictable.

This is bad for so many reasons: frantic people don’t do careful experiments or do good science, very smart people opt out all together and figure out how to make you click on ads instead, and a lot of time, money, and effort gets wasted in the churn.

If you want more cancer research and less ad optimization, nag your reps to improve how we fund research!


I feel like you could pull a 'natural flow' subset out.


I feel like searching for natural flow will not give me the results I want.


Not if you have a natural flow for all the random visits.


I mean, at that point you're having an arms race against Facebook's machine learning. My money's on them.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: