[hashberry]

I have a personal gmail and a work gmail (G Suite), each using my full name, but both run in separate browsers using the same IP address. I always wonder if Google keeps a "master record" of people and all their associations, if not for security or to work with law enforcement.

[s3r3nity]

>...the same IP address.

In my experience in adtech, not only can Google do this pretty accurately, but other third parties as well (e.g. DMPs and the like.) Even if they couldn't make a deterministic association, they have enough data points to make a probabilistic association with high likelihood (ex: "Given all these data points, we're 95% confident that these two people are the same. Therefore we're going to attribute the actions to the same person.")

Now, to qualify my response a bit, this isn't necessarily for security / law enforcement, but mainly for better targeting parameters. Example: frequency capping of ads (buyer specifies that you only see an ad X number of times in a given time period) or more relevant targeting (you don't see completely different ads in different browsers as if you're from two non-overlapping demographic groups.)

[caymanjim]

Facebook does this to build shadow profiles of everyone, even if they are not Facebook users. I'm sure Google does similar things. They certainly could.

[junon]

The thing is, doing this is actually quite trivial. Graph databases are used in large part for this very usecase.

[summerlight]

IMO, that doesn't make a good trade off. The identity problem space is already super complicated due to ITP while the risk is very high. If you make mistakes, you will accidentally leak personal information to others while the benefits... are pretty minimal, probably having 1~2% more revenue from those people with dual accounts. Given that joining signed-in and signed-out identity is explicitly prohibited as a condition of Double Click acquisition, I doubt if there's any incentive to do this kind of joining for Google.

[Romanulus]

Probably. I recall Project Veritas showing hidden video of a Twitter dev speaking about databases that link all profiles of someone (using things like IPs, fingerprints, etc. to stitch the relationships together).

[s3r3nity]

You're probably getting downvoted for discussing Project Veritas (which is a silly reason for downvoting IMO) but your point is still accurate: these companies are defnitely running models to create probabilistic identity models in the absence of deterministic associations.

Deterministic = I specifically say I'm person X and am logged-in. Probabilistic = I am not logged-in on this browser, but am on the same computer, same IP, and am logged-in on a separate browser at the exact same time under name X. Therefore I'm very likely person X.

[scohesc]

I really hate changing the topic here, but I feel it has to be said sometimes.

It's _very_ frustrating when people immediately discredit someone or something because they don't agree with them, even if they don't have the full story.

Project Veritas does some good work, even if they're blasted in popular culture/mainstream media for being 'biased', 'alt-right', 'etc'. I bet you the downvotes you're getting are just because you mentioned Project Veritas.

They have some very out-landish views, but when they actually put people on the street or go undercover, they've revealed dirt on a lot of companies and people.

It's shocking to me to see society go from "let's look at ALL sides of the coin, no matter how egregiously offensive they are to me" to "fuck 'em, they're trash media, they suck, they shill and are racist, alt-right losers and I'm not going to look at anything they post because in my mind everything they do is bad!"

I don't even go to Project Veritas outside of what I hear in the media, but I still give it a fair look and make my own opinion.

[logicslave]

Definitely. They have logic that links data sets, all it takes is one slip up for them to know youre the same person, probably with levels of probability about whether you are the same person.