I just don't buy this theory until I see some evidence. It wouldn't be hard for a talented individual to reverse engineer the Facebook client app to find out if it's either a) continuously recording and uploading raw audio to their servers, or b) continuously recording and locally processing audio before sending information about the content to their servers. Until I see an analysis of their code that indicates either of those things is happening, I have to believe that this is a conspiracy theory perpetuated by people posting these suspicious anecdotes all over the internet.

I had it happen to me too. Why would I lie about something like this? This is my real account, I'm a real guy.

My wife and I were talking about celebrating my kids birthday party at Arcade Foobar. A day later she started seeing ads on Facebook about Arcade Foobar.

It's grotesque and one of the many reasons why my wife and I no longer have facebook accounts.

Probably because people in your area often celebrate birthdays there. You would've seen it without the conversation as well. And it was one of probably 100 ads you and your wife saw over the days, you didn't notice the ones that were targeted badly.

You don't have to be lying about it, but consider that we often incorrectly detect patterns where none exist.

Two words, "lookalike modelling", or as FB calls it Lookalike Audiences:

https://www.facebook.com/business/help/164749007013531

Arcade Foobar want to draw in new customers, and so generate a lookalike audience based on data they have from existing customers. The modelling algo then compares data points for people who 'look like' their customers, and those people can then be targeted with a high probability that they will also be interested in the same thing.

This is a perfectly plausible explanation that doesn't require FB to be listening to you through your phone mic.

Complete nonsense. That you actually believe this instead of the much more plausible explanations are beyond me.

Why is it not plausible?

Because if Facebook was actually listening to your mic we would know it. They wouldn't be able to hide and it would be a massive scandal. It's simply too conspiratorial to take seriously when there is no evidence other than anecdotal stories like this.

Is this non-anecdotal enough for you? I'm tired of using the "conspiracy" label to shut down valid discussion.

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=786909

https://www.cnet.com/news/samsungs-warning-our-smart-tvs-rec...

https://www.washingtonpost.com/news/the-switch/wp/2017/03/09...

You don't see any difference between Alexa (which wouldn't work without listening to everything) and Facebook secretly listening without your consent to use it for ad targeting? None of your links provide any evidence this would be the case, or even mention Facebook.

The tools are there for you to try it:

https://www.charlesproxy.com/documentation/faqs/using-charle...

http://docs.telerik.com/fiddler/Configure-Fiddler/Tasks/Conf...

You could probably also do it with the emulator. You may need to do something like https://www.bettercap.org/blog/sslstripping-and-hsts-bypass/ or it could be even harder: http://docs.mitmproxy.org/en/stable/certinstall.html#certifi...

There are of course ways around this: https://serializethoughts.com/2016/08/18/bypassing-ssl-pinni...

If you need some more hints: https://github.com/ac-pm/SSLUnpinning_Xposed

https://forum.xda-developers.com/showthread.php?t=3034811

https://eaton-works.com/2016/07/31/reverse-engineering-and-r...

If you do this, please write it up. I'm sure that many people would be interested in this research if it hasn't already been done.

EDIT: In the spirit of POC||GTFO, I got curious and did it myself. I don't see any evidence of any mic uploads. The worst thing I see is it posting your current location. Here's a screenshot of the typical stuff I was seeing (this particular one is for the messenger portion of the app, and it looks like it's pulling in assets for stickers. I have the Pusheen pack because it's hella adorbs): https://i.imgur.com/dVekldx.png

It's POSTing to graph.facebook.com as you scroll to get more data. It's just JSON and pretty easy to figure out what's going on. I just followed http://docs.telerik.com/fiddler/Configure-Fiddler/Tasks/Conf... . I guess they're not pinning their certs, at least on my Nexus 7 and the version of the FB app that I used (v26.0.0.22.16 on Android 5.1.1 on the 2012 Nexus 7, since that's what I had laying around to mess with easily).

This literally took about 5 minutes to set up and get working. I encourage those curious to try it and see for themselves. I just read http://www.portero.com/blog/a-brief-history-of-gucci-infogra... aloud (I never shop for Gucci handbags, so seeing ads for them would be the tell) to test. I'm not some forensics expert but again, looks like nothing I wouldn't expect, definitely not uploading audio streams to Facebook as I'm speaking or +/- 5 minutes. Again though, check it out for yourself.

EDIT 2: I searched for a few unique strings from some of the decoded JSON my N7 was sending to FB and found https://github.com/macliu010/android_facebook_decomplied , which should be interesting for those following along at home.

While I applaud any attempt to gather real and first-person observations...

> I don't see any evidence of any mic uploads.

... I am somewhat concerned at what seems to be a common assumption that Facebook (or any other remote server) is sending the same software to everybody. I find that unlikely when "AB testing" and similar practices are are commonly used. Facebook is very strongly involved with "targeted" services which suggests they are more likely than most to send different versions to different audiences. Obviously that isn't proof, but it's a good reason to extrapolating from a single example.

Version numbers and hashes of the installer or binaries from all parties would help a lot.

> I encourage those curious to try it and see for themselves

That's always good advice.

If we're talking about mobile apps, it seems more likely that they would send the same software to everyone and just turn on these features being tested by sending some activation signal from a central coordinating service, or perhaps including it in some web content that the app might use. That seems much easier than actually distributing different apps.

> more likely

My point is that guessing without actual data isn't evidence. It's easy to speculate and guess at what we think is going on, but it's a bad idea to assume any particular speculation is right in the best situations. It's foolish when the subject of the investigation's core business model is based on sending different things to specifically targeted groups.

> just turn on these features

Perhaps. There are many ways a feature like this could be implemented. See the various obfuscated code contests for extreme examples of how functionality can be hidden in a program. This just means network logs from different clients are also needed, in addition to hashes of binaries.

How would you know if it is tunneling out the data, concealed as any other kind of traffic, and at some future point after recording?

I am pretty sure it would be undetectable if done right, given no access to the source code of the multiple Facebook components running on your device (some pre-installed by the manufacturer and unremoveable).

This would make an interesting blog post.

Thanks. I don't keep a blog anymore or I'd do it. They have also recently denied doing this: http://www.businessinsider.com/is-facebooks-app-spying-on-me... so if anyone can show evidence otherwise, it would likely be A Big Deal.

EDIT: Sounds like they might pin certs on newer versions of the iOS app. https://github.com/Naituw/HackingFacebook might be useful to get around that. I haven't tried it.

Should be reasonably easy to check with your favorite network monitoring tool.

It embarrasses me to see such conspiratorial nonsense on this website.

There are many reasons why such things happen: coincidence, someone google the topic from the same router IP address, ad tracking on Googled websites, etc etc

The idea that one of the worlds largest companies would risk their entire business by secretly recording their users for ad revenue is absurd on almost every level.

Yet some people on HackerNew, a forum that self-selects to a highly educated/intelligent part of the community actually believes it.

It's mind-boggling.

> The idea that one of the worlds largest companies would risk their entire business by secretly recording their users for ad revenue is absurd on almost every level.

Right[1]. Nobody makes risky[2], ethically challenged[3] decisions[4] in business[5]. Ever[6]. Doesn't happen[7].

[1] https://www.recode.net/2017/5/11/15628924/alphabet-waymo-law...

[2] https://www.yahoo.com/news/hampton-creek-serves-another-vega...

[3] http://fortune.com/2016/07/08/rise-fall-elizabeth-holmes-the...

[4] https://www.inc.com/business-insider/inside-lending-club-sca...

[5] http://fortune.com/2016/08/14/fraud-allegations-hud-skully/

[6] https://www.nytimes.com/2016/09/01/technology/a-silicon-vall...

[7] http://www.businessinsider.com/fbi-us-attorney-could-look-at...

> The idea that one of the worlds largest companies would risk their entire business by secretly recording their users for ad revenue is absurd on almost every level.

I'm with you on the skepticism, but they are one of the largest companies in the world because of ad revenue. It's not putting their business at risk, it is their business.

It embarrasses me to see such horrible logic and head-in-the-sand thinking on this website.

Whether they record ambient audio or not, how is it implausible or out of character with everything else they do? Facebook and many other companies collect countless dimensions of data on users to profile them for marketing. Audio is just one more. It's not even really that disturbing or far fetched. Why would they ignore such a lucrative data source when they've already tapped so many others?

It's a direct business plan with Amazon's echo for christ sake. What is so hard to comprehend about this?

Not wearing my tin foil hat, just reporting what I found when I tried to test it myself. Only our computers on the router, had never googled the topics.

See my updated post just above. I just tried it with my Nexus 7 and saw nothing around mic uploads. Install Fiddler and MITM your Facebook app and repeat your experiment. I'm very curious to see if you find something different from what I found.

My wife and I did the same test with the same results. It's the reason I took the app off my phone and just use the mobile web.

For clarification on what we did for our test, I live in the south east. We don't watch professional baseball at all...so we started talking about going to a Cincinnati Reds game.

And then we saw ads for the Reds and MLB.

I'm not trying to go full tin foil hat here, but it was a pretty strong coincidence for that to happen.

Did you also see ads for things you didn't talk about? How did you select those very obscure topics?

are you serious? Which app/phone?

We both had our phones (one iPhone, one android) and macbooks on with facebook (web/app) open.

With facebook web it's pretty easy to say as you have to give permission via your browser. Otherwise, for Chrome, you'd have to assume that Google works with Facebook to help them make more money. Rather unlikely.

How did you ensure that no other apps were listening in the background?

I don't understand--how would other apps trigger related facebook ads?

Send some data to an ad network or something that feeds it, Facebook ads somehow pick that up. Not beyond the realm of feasibility. Try using a just-factory-reset phone with JUST the Facebook app installed. There is science to be done here.