Comment by ghostwords | Hacker News Reader

Comment by ghostwords | Hacker News Reader

Parent Full thread

Hi, Privacy Badger dev here.

While there is likely to be overlap between the various advertising/tracker lists and Privacy Badger, regardless of whether you enable learning locally, Privacy Badger can automatically discover new trackers that list-based blockers don’t know about.

Besides automatic tracker blocking, Privacy Badger comes with privacy features like click-to-activate replacements for potentially useful trackers (video players, comments sections like Disqus, etc.), and link cleaning on Facebook and Google.

Privacy Badger is also a political tool. By using Privacy Badger, you support the Electronic Frontier Foundation [1]. Privacy Badger sends the Global Privacy Control [2] signal to opt you out of data sharing and selling, and the Do Not Track [3] signal to tell companies not to track you. If trackers ignore your wishes, Privacy Badger will learn to block them. The idea isn't to block all advertising but rather to promote a better Web.

[1] https://www.eff.org/ [2] https://globalprivacycontrol.org/ [3] https://www.eff.org/issues/do-not-track

Wasn't the local learning disabled because it enabled fingerprinting?[0] Has that been fixed since you're recommending it?

[0]: https://www.eff.org/deeplinks/2020/10/privacy-badger-changin...

I'm not sure what you are asking. What am I recommending?

All extensions that modify page state are fingerprintable. Privacy Badger's local learning creates the possibility of additional fingerprinting. As discussed in the blog post you linked to, we decided that we can deliver most of the benefits of automatic learning without local learning. We further believe that local learning can offer benefits that outweigh the risk of additional fingerprinting, which is why we kept the ability to re-enable this feature at the user's discretion. Does this answer your question?

> We further believe that local learning can offer benefits that outweigh the risk of additional fingerprinting,

Nice. I thought the local learning (which was kinda the main appeal of PB) was dead after reading that post. Glad that isn't the case.

We've been trying to communicate that nothing essential changed, Privacy Badger still learns automatically, but we clearly have a ways to go before this becomes generally understood.

Thanks for your work.

Has there been any study as to how closely Privacy Badger converges to list-based blockers like uBlock Origin given enough time? Is that at all a metric you use in development?

Definitely, we have been working on generating insights from comparisons to other content blockers as part of Badger Sett scans.

There are a few academic studies that note that algorithmic and manual approaches to tracker blocking tend to compliment each other, for example:

https://kevin.borgolte.me/files/pdf/www2020-privacy-extensio...

Thanks!

Do any significant disadvantages of Privacy Badger stand out to you, contrasted with uBlock Origin? The biggest problem I can think of is that uBlock Origin will immediately prevent me from loading content that could be dangerous, while Privacy Badger has to visit dangerous content enough times to learn to block it, which is less good for security.

On the other hand the big advantage to me for Privacy Badger is that it doesn't depend on manually-curated lists. uBlock Origin nicely wraps the block lists but when it comes down to it you are giving control over your browsing experience to the random volunteers building these lists, which, who knows? Though I'm not accusing anyone of anything.

It may be that Privacy Badger is more likely to break the Web than manual list-based blockers, just due to the nature of automated learning.

Regarding having to see dangerous content multiple times:

- Privacy Badger gets remote learning at this point from Badger Sett, so most common trackers should be accounted for.

- Privacy Badger isn't so much a security tool as a privacy tool. Yes, blocking trackers is good for security, but it's not the primary objective. The primary objective is to stop non-consensual pervasive tracking. In this light, having to see tracking a few times before deciding to block it is a reasonable approach.

Reply on news.ycombinator.com