Pretty sure they were breaking all kinds of PII laws.
I really don’t understand the goal with vague statements like this that can’t even provide even the slightest hint of specifics.
What specific data? Even a single example would make this anecdote useful. Instead it feels more like a brag. “I know something but I’m not telling” but in this case the commenter doesn’t claim to have worked at Facebook (just the industry in general) so I suspect it’s hearsay anyway.
> Pretty sure they were breaking all kinds of PII laws.
Given the way Facebook has been under the microscope and dragged in front of Congress, I’m going to assume that their corporate counsel was very careful to provide at least a best-effort attempt to comply with every law available at the time. It may not be popular, but I really doubt Facebook was violating laws for a decade straight as the largest player in the space.
You can quickly create an account and look at their self-serve ads. There's no reason why anyone needs to try and "guess" what these tracking tools can do. You can just go to the endpoint of that collected data and just see what you can target.
It's better now but before you could do EVEN more. But better in the sense that someone who needs a limb amputating to stop gangrene setting in is better.
The ubiquity of user tracking is extremely useful yet culturally absurd. Now that’s ambiguous ;)
It’s changed a bit recently with GDPR, the Cambridge Analytica scandal and some third party cookie privacy stuff so it’s a bit less insidious now, but it’s still pretty bad.
Frankly I do not know why corporations don't block facebook as a security risk. Seriously that stuff is bleeding info on what your employees are up to.
Key points: https://twitter.com/YBenkler/status/1070337233159372806?s=20...
Here was one thread of highlights which is still somewhat readable: https://web.archive.org/web/20181206132832/https://twitter.c...
Germany banned their cross-site data sharing/reciprocity, such as from menstrual cycle-tracking apps which had come to light (maybe separately), in 2019: https://twitter.com/YBenkler/status/1093495901342126080?s=20...
The difference when it comes to other industries (e.g. food) is that the regulation has had time to develop, and most legislators understand the concepts. So it's harder to cheat.
Forgetting someones opt-out preferences by mistake doesn't ring as severe as using light carcinogens in your food mix.
I'm pretty sure everyone of technical aptitude knew Facebook's data faucet. But maybe I missed something.
As far as I know, Facebook:
a) Had all the freely provided data, PII/likes/social graph/etc.
b) On Facebook's site or mobile app, the were fingerprinting your device, examining your scrolling/mouse/clicking/other inputs to determine attention on a page
c) Could recreate most nonuser's social graphs just by seeing them as endpoints in registered people's contacts c) On the web, had "like" buttons or ads and their code pretty much everywhere. Therefore they could track most people to most sites.
d.1) Sites could directly provide more information to allow retargeting d.2) Sites could directly provide more information through a host of other services FB offered to the developers
e) On mobile, in the background, scrapped your contacts, GPS, nearby devices (other phones, WiFi, Bluetooth, although tower information may or not be included). Also, had installed by default on a lot of phones
f) On mobile, provided libraries people could import into their apps, mostly but not exclusively for ads. This let them get similar insights into usage patterns as on the web. Also, if people didn't install the FB app, let them get (e)
g) Used your real identity to purchase information about you from the various realworld data merchants
Which did I miss?
I don't think you missed anything substantial... but I'll add two extensions:
(1) Social graphs change slowly. And they still own Instagram, so for many users they have a live social graph still.
(2) Facebook Pixel is dead. Long live Facebook Pixel!
The Facebook Pixel is now (or at least nearing) effectively dead on modern up-to-date devices running ad blockers. Of course that leaves plenty of desktop machines where people aren't running ad blockers.
But more importantly, Facebook has acknowledged the elephant in the room and moved from client side to server side with Conversions API (CAPI) (aka the new "Facebook Pixel"). And there's nothing ad blockers can do about server-side analytics...
Imagine gloating and being proud of such a career.
In a completely different matter, I worked on high level space programs in the late 1970s and if you had any idea about the information that the government is hiding on extraterrestrial life TO THIS DAY, your hair would stand.
I doubt it's on FB during that period though?
I would guess though that a bunch of health tech sent (perhaps accidentally just not understanding) a bunch of patient data though. Seems they are the responsible party.
There's been other examples beyond FB of 'auto track' too. devs just don't know or forget to turn it off.
Not to mention for some reason at that time putting a FB like button on all the porn. Who clicks that?!?!
I worked for a "healthtech" company in London at the beginning of the pandemic. They had the Facebook SDK malware embedded in the app that people were supposed to use for GP consultations.
I don't believe any explicit health data was sent (there was no intent to do so, and I’m not sure if that would even be possible), but merely the fact that I'm talking to a doctor (and the current time, location, device fingerprint, etc) is not something I'd like Facebook to know.
I know breaching the GDPR is basically the norm in any tech company but I thought that being involved in healthcare would make them super risk-averse and make an extra effort to comply.
They were not alone in this - PatientAccess and a bunch of other sites - that you can use to book GP consultations (including through the NHS - UK’s socialised healthcare system) had a shit ton of such trackers too, obviously loaded before any GDPR consent could even be obtained.
It was a straightforward call to get info about the user, including name, email, interests, etc. Their friends list. Info about all their friends, including all the same details. And so on.
There was a EULA where the developer had to promise to delete all the info when the user signed out of the game, and not to share the info. That was the only security.
The project fizzled, but when the Cambridge Analytica news broke, it confused me, because my recollection is EVERYBODY had all those lists of user info. Seriously, tens of thousands of different companies, with the only thing stopping them was a pinky promise.
From memory this has now all been rolled up into cohort demographics and 'look-a-like' audiences so you can no longer break your data down by specific users demographic attributes or target ads by say an email list unless they are already your users (and it's used for specific types of ads; retargeting).
From memory some of the more unsettling breakdown/targets were
- Ethnicity
- Life events
- Politics
- Pages (so other businesses) they had liked
Now they just use the facebook pixel to put you into look a like audiences because they can see you went to the home insurance website and the real estate website (they all have the pixel installed) and make the same broad assumptions we were previously making but on the cohort and not your specific unique identifier.
Why shouldn't I be able to advertise to people who have recently marked themselves as married, or who liked the Yankees facebook page?
You also shouldn't be able to upload a list of email addresses, target your ads to them, and then use Facebook's analytics to see how many of those people have divorces or an investment property via the segmentation analysis. Depending on how small that list is, a lot of that data starts getting very specific to an individual.
Facebook obviously also thought you shouldn't be able to do this since now you can't. Everything is now cohort and look-alikes.
Additionally your acting like Facebook is only putting you in the ‘recently married’ bucket if you marked yourself as married. Facebook is smarter than that, they are putting you in these buckets based on your messages, instagram activity, and browsing behaviour, not necessarily based on public information you expose in your profile.
The above, is all frowned up (for the record, I 100% agree) what is interesting is, our civilization has evolved to frown up the above "categorizing" of people, but from that list (only ethnicity) is not "choose able" by the person.
I.e I can mostly choose when I get married and to whom and which party I support but I def cannot choose my race. Yet the above is all considered 'equally bad'.
Just something I noticed.
'offline conversion' still allows you to send in names, age, bday, gender, etc for matching, IP, UA.
Though now it's hashed before going to FB.
And you can pass almost whatever custom data you want in. So I can in my industry optimize for a long term political donor, or potentially an early vote. Or someone accidentally sends in 'this person bought hepatitis meds'
This still exists despite what someone below said, unless I'm totally missing or misunderstanding something.
However it is not as valuable, and shrinking audience able to do deliver to because of iOS restrictions. And likely with Chrome too eventually
FB used to have more detailed interest/demographic buckets to target that they supplied. Used to be able to type in basically anything from what someone likes ___ very niche page to engages liberal political. There are still interests but there are fewer of those 'sensitive' ones. Still lots of stuff like works at ___. of course age, gender, geo.
But more fine grained interest targeting seems like going away pretty soon it's just going to be broad demographics.
The ROI is just not as good without iOS fine grained targeting FB is having to do a bunch of tricks with AI/modeling to try and make it perform but it's not as good.
I'm for targeted advertising. I think what iOS is doing is uncompetitive and bad.
But I do think there should be some sensible regulation. Like no healthcare or sensitive topic data (LGBTQ, dating, etc).
** ADDITION sorry this is long but one additional thing I think people also confusing the ad product with the old FB api.
The old FB api was an absolute sieve you could get basically any data a person has on their profile and also their friend's data. This is what happened with Cambridge Analytics.
All that has been shut down even login with fb they are way more strict about actually testing sites etc
It is however an easily-reversible hash, by design as that's how FB can correlate between the different datasets. When it comes to finite sets such as phone numbers or dates of birth it's also trivial to search the entire space by bruteforce.
For sure on a rainbow table for something like cell phone. but i don't know why that would matter? Anyone can generate all the possible phone numbers.
The whole point is that it is matchable. Like if they already have my email then they know if it's a match, but if they don't have my email they don't know what the missing email is.
Like what's my email from this (without knowing my email) below: 2c03e4a168bed89f5208250cdefbe97d4d87ba7812df896311676acc2ddfcdb4
Names and emails can be bruteforced with various lists from existing data breaches or data brokers and you'll probably reverse 80% of them.
However reversing them is not even necessary - an adversary like Facebook can infer it based on other data, for example, let's say they know your phone number but not your email - now you buy/sign up to vendors providing both that phone number and email and they provide it to Facebook - now Facebook knows that you signed up to those vendors with your number (as they have the plain text value, can hash it on their side and compare), but they also see that there's a mysterious email hash - they don't know its plaintext value, but it perfectly matches the same vendors that have your phone number. They can infer that it's probably your email address, and while they still don't know what it is, they can use the hashed value to track you across other vendors without ever having to know the plaintext value.
And no doubt it's not compatible with their current business model. Which is the point, it's a model that exploits property that isn't theirs for unfair gain.