Hacker News Reader

Comment by saddlerustle | Hacker News Reader

Lots of people in this thread are claiming that it doesn't based on gut feeling, but consider:

- Facebook publicly claims it does [1]

- Mark Zuckerberg testified in-front of congress and stated they do [2]

- Multiple government regulators have specifically checked that it does in their privacy audits. [3]

They might be lying and actively conspiring to not delete it, but they'd have to have a very good reason to take on that much legal risk.

Now, consider:

- A infinitesimal fraction of facebook users try to delete anything.

- Facebook makes money from your data by showing you ads. If you stop using facebook, you stop generating any revenue and your data becomes a liability, not an asset.

What reason would there be to lie about it?

[1] https://facebook.com/help/250563911970368

[2] https://www.washingtonpost.com/news/the-switch/wp/2018/04/10...

[3] https://www.pdpjournals.com/docs/87980.pdf

View on HN

> Mark Zuckerberg testified in-front of congress and stated they do

I just want to point out that "testified in-front of Congress" seems to have no relation to the truth of any statement. I don't recall the last time (if ever) there have been legal ramifications for lying in front of Congress.

Zuckerberg also testified that shadow profiles don't exist and that users can always remove their personal information from Facebook (they do exist, and you can't). James Clapper testified that the NSA doesn't "wittingly" surveil hundreds of millions of Americans (they do, very "wittingly").

> I don't recall the last time (if ever) there have been legal ramifications for lying in front of Congress.

It's one of the reasons Cohen is in prison, isn't it?

He lied to the FBI after a plea bargain which involved him telling them the truth. From memory, this all happened before his Congress testimony but it's definitely not related.

It is illegal to lie to Congress (unrelated to whether you're under oath or not), but I can think of very few examples where there were real ramifications for it.

> He lied to the FBI after a plea bargain which involved him telling them the truth.

Sure you're not thinking of Manafort? I don't think Cohen had a plea deal, and I don't think he was charged with lying to the FBI. Manafort and Flynn did, and Manafort I think was the one that then broke the deal. There's a lot of crimes to keep track of.

I guess it depends on what you mean by real ramifications as it's being served concurrently but Cohen is serving two months for lying to Congress, a charge he plead guilty to.

Yep, that looks right from his Wikipedia article. One count of making false statements to a congressional committee. It’s not immediately obvious to me when he was charged/convicted of that though. https://en.m.wikipedia.org/wiki/Michael_Cohen_(lawyer)

Cohen pleaded guilty to several charges [1], including some that alleged that he lied in a written statement sent to the Senate Select Committee on Intelligence and House of Representatives Permanent Select Committee on Intelligence on 28 August 2017, and then repeated those lies in his testimony to the Senate committee on 25 October 2017.

[1] https://www.justice.gov/file/1115596/download

Testifying in Congress absolutely carries legal ramifications. It's just pretty hard to pin it so I agree with the "difficult" sentiment.

https://techcrunch.com/2018/04/11/is-zuckerberg-under-oath-c...

John M Poindexter was one who was convicted of lying to Congress: https://en.wikipedia.org/wiki/John_Poindexter#U.S._executive...

I agree that it is technically illegal, but if the most recent example of a conviction (that you can think of) was almost 30 years ago that tells you that it's effectively unenforced.

Michael Cohen was sentenced for exactly this [1] just a few months ago.

[1] https://www.economist.com/united-states/2018/12/13/michael-c...

Yet James Clapper somehow got away with it. As do a lot of other people who have the same "political opinions" as Zuck.

Didn't that also involve lying to federal investigators?

Does it? One conviction in 30 years is also consistent with the conclusion that not that many people get invited to give evidence to Congress, and the people that do tell the truth (or at least don't tell provable lies).

Definitely illegal, not just technical.

Was the nature of deletion explicitly defined as "deleted records are removed from all databases, data warehouses included"?

They technically wouldn't be lying if they only deleted the record from application systems, but retained it in historical data by simply setting a bit flag (eg "IsDeleted") to 1.

The question about what motive they have comes down to whether or not the types of posts a person deletes indicate something meaningful about their personality. Ad analytics are ultimately seeking to understand the kind of person you are, after all.

I read the audit, and here's something I noticed (other than that it's 8 years old):

"In determining appropriate retention periods for personal information, data controllers can have due regard to any statutory obligations to retain data. However, if the purpose for which the information was obtained has ceased and the personal information is no longer required for that purpose, the data must be deleted or disposed of in a secure manner. Full and irrevocable anonymisation would achieve the same objective." (Page 69, under 3.4 Retention)

So basically, as long as it's anonymised, the data can be retained.

But “if the purpose for which the information was obtained” was to enable better ad targeting, then “data retention” is still “appropriate”, eh?

Under this reading, is there even an obligation to anonymize? And how anonymous does anonymous need to be? Does a simple base64 encoding count as “anonymized”?

I had that thought originally too - however the keywords here are "statutory obligation" (basically, legal reasons.) So if the FBI asked them to retain the data, they can do so as long as the FBI needs it.

And the matter of anonymisation is a great question - "irrevocable" anonymisation certainly has a different meaning back in 2011, when swapping a name for a guid would do the job. Nowadays, it would require at least deleting all relationships as well, since social network analysis is much more advanced these days (especially at FB, of all places.) It wouldn't be impossible to derive the identity of an anonymous account/record based on the undeleted data associated with it. And since FB's "ghost profiles" are something we know exist, I think it's safe to assume those relationships are being maintained somehow.

<cynical interpretation> The data can be retained so long as Facebook is in the business of targeted advertising, and that's "the purpose for which the information was obtained".

Also, I'm now looking forward to Zuck's next apology for "a breach of trust" and his explanation of how they "failed to live up to their own standards" when it becomes publicly undeniable that their "Full and irrevocable anonymisation" of data they've claimed is deleted is as flawed as all the other attempts we've seen of doing that.

> eg "is_deleted"

FTFY

Facebook don't need to keep your past posts, their algorithms have already run the data. The real question is whether you can reset the "personas" they assign to you....

Agree. If it is deep learning do not know any way to get from weights back to data.

There is some research being done on this topic. At least in ways to sign models so that you can test copyright based on certain input keys.

You can't un-train the model. Your contribution has already been made.

Either

#1 You can train the model continously

#2 the model stops learning

#3 it has a submodel which detects you are feeding it false data.

Given that people change over time, #2 seems unlikely. The question to me is "how advanced is #3?". Another thing I wonder about is how relative are these to eaxh other. At odds?

#1 is a well-established technique called "online learning", and I'd bet money that it's how many of these "industrial scale" ML algorithms are trained.

#2 makes no sense from a business perspective.

#3 is also well-established. This is how Google's Captchas work, for example.

Instead of deleting posts or garbling them, one could replace them and add new posts that portray a false persona.

Check out an open source browser extension called track me not that does this with search engine queries

There are ways to be reclassified as a undesirable persona

> "Facebook makes money from your data by showing you ads. If you stop using facebook, you stop generating any revenue and your data becomes a liability, not an asset."

If my friends or acquaintances continue to use facebook, then information about me continues to be useful to facebook as that information can be used to complete their model of what sort of people my friends are, that they would associate with somebody like me.

I wouldn't trust Zuckerberg to pour a glass of water on a burning orphan, let alone delete data for real. Nor would I trust any politician to know anything about the subject.

Maybe not lie per se but they might have multiple data stores, one for the "social media" side of things and an archive for machine learning, one for targeted ads, etc. So they could legitimately say "we delete your data [from the perspective of social media]" while still having an archive for their other business units.

It wouldn't be the first time Facebook (nor any other business that profits from people's data) have been deceptive while treading a careful line between honesty and lies.

edit: I should add that if they were keeping archived records in a separate business unit, there is the possibility that they keep the original or even all edits. So scrambling your data might not do much aside confusing those who have you in their feed.

Under GDPR, you cannot tell users you deleted data if it's still available for other purposes (except for legal requirements like invoices).

Given FB's size and the small amount of people deleting data, I doubt Facebook would take the risk to keep a few more datasets.

Facebook looks to be willingly not following most of GDPR https://ruben.verborgh.org/facebook/

The UI doesn't say "delete", it says "remove". Remove could mean "hide from view" or it could mean "delete the underlying database entry". It's a bit of grey area where it could be argued that users have consented to by using Facebook.

Granted GDPR is supposed to catch businesses that pull those kind of stunts, you have to remember that Facebook do already break GDPR in number of public ways too. So it's pretty clear they have a relatively open interpretation of the regulations (and an army of lawyers who are confident they can proceed in such a way). Or it might just be the case that even the worst fine issued by the GDPR is worth the risk given the financial benefits awarded to Facebook for retaining data.

Facebook has also states it will not conform to the gdpr outside of Europe

>- Facebook makes money from your data by showing you ads. If you stop using facebook, you stop generating any revenue and your data becomes a liability, not an asset.

This part at least is incorrect. Your data will always be an asset, and FB makes money not just by showing you ads but by selling your data.

How does Facebook make money selling data, specifically as it's different than selling ads? I thought their revenue stream was 100% ad-driven.

The corpus of all of your posts are valuable for statistical analysis for ML models, sentiment analysis, and the like.

Isn’t that all to sell you ads?

For the general "you" yes, but not the specific "you". It would be possible to use your data to sell someone else ads.

But they don't? They use data internally to improve ad revenue but I doubt they ever sold data for money or intend to do so.

Do you know what the legal status is of models derived from someone else's data? EG, can I train an AI from every Marvel movie, then use that to produce other movies?

That's not a great analogy because I would be violating other people's IP, whereas FB generally has rights to the data they collect.

They obviously exchange (precisely targeted) attention for money: https://www.facebook.com/adsmanager/creation

And they obviously give away some data for free, especially pertaining to users who click through OAuth consent screens as in the Cambridge Analytica case: https://developers.facebook.com/docs/graph-api

What are some examples of transactions where they exchanged data for money?

Facebook cares about “your” data, but that doesn’t really mean your painstakingly edited review of the new avengers movie. It’s more what you spend time looking at, what you click on, etc.

It's perhaps likely that Facebook doesn't still hold the data, but a 'partner' keeps it for marketing/research purposes...

Facebook has no incentive to provide access to third parties. This data is their source of revenue; they sell services off of it.

Except that they have repeatedly had gaping holes in their security that just happened to result in mass ex filtration...

It would’ve surprising if they didn’t capitalize on them.

https://en.wikipedia.org/wiki/Cambridge_Analytica

Whoa, an encyclopedia site anyone can edit!

They did not sell information to Cambridge Analytica. As stated in the page you linked, the data was scraped using a 3rd party application requiring user opt-in. Now, there was a bug that allowed the app developer to also scrape data from the friends of the actual users but that has been patched and the developer's license has been revoked as the entire ordeal was against FB's licensing agreement.

I am not a fan of Facebook but this entire thread is just filled with misinformation.

Pretty much exactly this, but also:

I work at a company you've heard of, and you've probably (P > .5) used at least one of our products. We're international (including EU). I personally wrote our user data deletion logic for GDPR compliance.

We delete all of your PII. It doesn't matter if you're from the EU or not, because it's too expensive to figure it out and too risky because you're going to miss some weird edge case - Legal doesn't have much of a sense of humor when it comes to wiggle room.

My experience has been that larger organizations consistently spend more effort (relative to size) to genuinely comply with privacy regulations than smaller ones. The risk:reward ratio for deliberately ignoring or subverting privacy regulations is insanely bad. There are too many surfaces along which that would leak out, and the gains would be pretty marginal. Pretty much anyone who works at a large tech company can and will confirm that this is the case (see Jeff Kaufman's posts on this thread). There is no conspiracy between the six-digit number of engineers who work at these companies to keep quiet about it.

>There is no conspiracy between the six-digit number of engineers who work at these companies to keep quiet about it.

I wouldn't say "conspiracy" but they've been mums-the-word about things like shadow-profiles. Perhaps, it comes down to too much kool-aid but to say that engineers would initially speak-up in such cases has been proven wrong, time and again.

Take the Snowden revelations, as an example: How many years were the programs in service before Snowden went public? How long have we known about shadow profiles and no one from Facebook has come forward to say, "yes, this is what they're doing and it's wrong"?

Relying on people to do the good that should be done by the organisations doesn't take into consideration that those engineers face severe penalties for "going pubic" about such things - namely because whistle-blower laws do not supercede such thing as NDAs.

Doesn't Facebook still have a zoo of MySQL schemas in production for each table? While I'd believe the product works as intended in some fraction of production, I think it's reasonable to suspect some or all of the system is broken for a non-trivial set of users and/or groups.

And then, what about all the data copied into Hive? Copied to 3rd parties? Does Facebook go and delete data that 3rd parties took?

  >>> They "trust me"

  >>> Dumb fucks

https://en.wikiquote.org/wiki/Mark_Zuckerberg

You'll no doubt soon be scolded for mentioning that quote, but it will continue to be relevant for as long as Zuckerberg's actions continue display that mentality. We'll soon have somebody telling us he said that years ago and people change, but the only thing that changed with Zuck is he became a bit more diplomatic and guarded with his language. Actions speak louder than words and Zuckerberg's actions speak clearly.

He's the same old dirt bag he ever was. Here are some fresh stories to back that up:

https://www.theguardian.com/technology/2018/nov/21/facebook-...

https://www.vox.com/2018/11/14/18096008/facebook-zuckerberg-...

Yeah. If there was some evidence that Zuckerberg's overall approach had changed, I too would argue that continually resurfacing an old quote until the the end of time is unfair and unproductive.

All of Zuckerberg's and Facebook's actions, however, continue to suggest that his approach hasn't changed.

Yeah, for a funny example, we seem to have (mostly) stopped reminding Google of their old "don't be evil" slogan, since it became obvious their overall approach had changed ;-)

You got anything else? I am not really impressed by those stories.

I used to be one of the scoldiers saying things like "oh all young nerds say dumb stuff when bragging to their friends"

Honestly the vibe's still there if you pay him/fb enough attention. I'm working on extracting myself completely from them (Still got a few group messages on upcoming events to clear)

Not sure why you are downvoted - always good to remember. In many ways thats what he kept saying until last year, even to congress - just with nicer, politically more appropriate words.

It's always good to remember this and we should keep repeating it.

This is the most profound insight we have into the mind of Mark Zuckerberg and what he thinks of his user base.

Stop listening to the PR and examine his actions since then. You can see how that comment was not merely a youthful indiscretion, but his entire business.

Zuckerberg continues to display this exact same mentality, only now he has billions of dollars in resources.

One of the first thing I learned in the IT industry...do not delete any data no matter what the circumstance is! And I believe internet companies whose life and soul is data would not want to delete it

Simply archive it in an "emergency backup".

> - Facebook makes money from your data by showing you ads. If you stop using facebook, you stop generating any revenue and your data becomes a liability, not an asset.

I doubt this

> Facebook publicly claims it does Hahahahaha YMMD!

> Mark Zuckerberg testified in-front of congress and stated they do He would tell us everything that helps him as he is an opportunist

> Multiple government regulators have specifically checked that it does in their privacy audits You're talking about the governments secretly spying on us and lying about that since decades? Ouch.

There is no legal risk for those who observe if you abide the laws.

> A infinitesimal fraction of facebook users try to delete anything What does that tell you about the majority of ppl?

> Facebook makes money from your data by showing you ads. If you stop using facebook, you stop generating any revenue and your data becomes a liability, not an asset. That's just what you think you know. What about psychological profiling, law enforcement etc.?

Also did you hear about the "shadow profiles" about users who don't have an account? The moment you're surfing into their net (as we know even through embedded like buttons etc.) you'll be milked like a dairy cow.

> What reason would there be to lie about it? What reason is there not to be honest about it? The recent events have shown very clearly that they don't have to fear much.

> They might be lying and actively conspiring to not delete it, but they'd have to have a very good reason to take on that much legal risk.

Or they can just claim, again, that they accidentally forgot to delete copies of these posts from their backups. It just completely slipped their mind.

Mark Zuckerberg is a liar who has demonstrated that he will do and say whatever is in his and Facebook's best interests. If you believe anything other than that, you're a fool.

Reply on news.ycombinator.com