- the data that Facebook collected about you from various other sources, e.g., data brokers

Mr. Zuckerberg and other Facebook staff assert: "We do not sell data."

Does Facebook acquire/purchase data about people from other sources, e.g., data brokers, and then combine it with (a) the highly personal data people voluntarily submit to Facebook and (b) the highly personal behavioural data people consent to allow Facebook to collect from their use of the Facebook website, Facebook app, and websites hosting Facebook anchors.

Is Facebook allowing users to download all the data Facebook has on them?

PSA: Facebook looks at what websites you visit in an incognito window, provided that the incognito window is open the same time as a regular browser.

How does it do this?

Apparently, incognito mode uses the same temporary cookies folder as regular chrome, so FB looks at this folder if incognito is open in one window while FB is being used in another.

Are you sure about that? Wouldn't it mean that all third party trackers could continue to track you while in incognito mode? How could it be called incognito mode if sites can still read your non-incognito cookies?

I'm 95% confident. And yes, I think it does mean that. The trackers have to be active in a non-incognito window for it to work. It reads the shared temporary cookie folder.

link to proof of concept?

edit: tried it on chromium 66: doesn't work.

steps to reproduce:

1. open new (non-private) tab, go to http://httpbin.org/cookies/set?k1=v1&k2=v2

2. check http://httpbin.org/cookies/, the two cookies should be there, and if you check the dev console, the expiry date is 1970-01-01 - 1 second. firefox says it's "until end of session".

3. open new private tab, go to http://httpbin.org/cookies

4. cookies from above do not show up.

If incognito mode is that broken in Chrome, and still allows sites that are open in normal tabs to track you, that would be very newsworthy.

agreed

I had to know the answer, so I wrote a quick script to test it. It appears that incognito windows can't read cookies from normal windows. I've linked to the source code there. If I've missed anything, let me know.

https://cookiechecker.netlify.com/

So cool that you built that!

Here's what Google says about cookies in incognito mode: "If you use Chrome in incognito mode (or in guest mode on Chrome OS), it will not transmit any pre-existing cookies to sites that you visit. Sites may deposit new cookies on your system while you are in these modes; these cookies will only be temporarily stored and transmitted to sites while you remain in incognito / guest mode. They will be deleted when you close the browser, close all open incognito windows or exit guest mode." Source: https://www.google.com/chrome/browser/privacy/archive/201408...

I first noticed that FB might be able to detect sites you visit while in incognito mode when I noticed that I was personally being retargeted by sites I visited in incognito mode while on Facebook. Here's an interesting quora thread I found while looking into it: "Also the cookies act too important role here. Cookies during Incognito Mode are saved in temporary folder, and they get purged after session ends. But when browsing Facebook and Google in the same session, they both share same cookies folder and cookies help to send tailored ads to users." https://www.quora.com/Im-using-Incognito-mode-for-Google-sea...

Curious for your take.

"It appears that incognito window can't read from cookies from normal windows."

To be clear, the claim I'm making is that normal windows can read cookies from incognito windows.

A simple test suggests otherwise, using Gruez's recipe above but swapping the private and non-private windows.

There are potential information leaks, such as the combination of IP address, user agent, screen size, and fonts available. And there may be bugs in 3rd party plugins like Flash that fail to respect private mode. But it doesn't seem to be anything as simple as cookies being shared.

Thanks. If they're not using the temporary cookies folder, I'm curious how Facebook shows ads based on incognito browsing. Some form of device fingerprinting?

no it doesn't. https://news.ycombinator.com/item?id=16908185

Yes, it does. And your client side “steps to reproduce” do not, in fact, reproduce server-side client fingerprinting to correlate between two disconnected browser identities.