Countless questionable youtube videos, reports of all sorts by people who have no idea what it would entail to perform such a stunt on a large scale.
After googling extensively, I still can't find any real scientific proof that it's either true or on true. It's mostly about some people relating their funny experience and mostly about Facebook saying "No we don't"
Anyone familiar with Charles Proxy would know it's pretty easy to spy on the traffic going from an app to the backend. I work in Facebook advertisement and I've been doing it for various reasons already many times. I've never seen anything that would like a speech datagram, but then again, it wasn't what I was looking for when listening to the Facebook app.
I would expect that given Google and Apple should be scrupulous about what the Facebook app does and doesn't, there should be a host of people who've been looking into that urban legend a number of time. Also the data plan of many people would simply explode too quickly for this to be true.
With my colleagues, we've been wondering if it would be possible to rely on a local text-to-speech engine running on the phone. Needless to say that would drain the battery pretty quickly and the dataset needed in the app would be pretty huge... But is it? And would it really drain the battery? More than Pokemon or the Facebook app...?
Who among you guys is a seasoned user of Charles and could run the experiment a tad further?
I can't help but remember my incredulous reaction to ECHELON in the 90s and how it turned out to be not just true, but much worse than the original "conspiracy theory".
A very relevant example: https://jscholarship.library.jhu.edu/handle/1774.2/36569
The microphone privacy invasion rumor has spread so far without evidence because people are tired of the 1000 other privacy invasions for which there is evidence. They don't care whether or not this specific instance is true. People are tired of FB being openly scummy at every other turn. FB is the least respected tech company for a reason.
I do work closely with Facebook (For a Facebook Marketing Partner called Smartly.io) and we collaborate closely on certain aspects. I think I have a very deep understanding of what Facebook is capable of and that's the reason why I want to debunk this. Also, I am really tired of hearing the questionable reports.
Look me up if you which: https://www.linkedin.com/in/momeunier/
people are scrambling to debunk an ill-defined phenomenon that's only exists anecdotally... and in the realm of those who ask "qui bono?"
who benefits if "facebook is listening"? what are the incentives?
the question of technical possibility is the first place people will go to... without realizing that if there is a will, there is a way.
Native apps on a phone are not comparable to government spying on networks outside of your control. The attack vector in this case is entirely within the device in possession: the code, and interactions with the OS and hardware.
Entertaining the opposing view: If fb did build this, how is there no hard evidence of it? On iOS, their code is distributed via Apple, so they couldn’t doctor versions for specific users, and any attempt to build self modifying or self deleting mechanisms would be detected sooner or later for similar reasons.
Nobody seems to be doubting the feasibility of building such spyware but rather are asking where’s the hard evidence and how could they get away with it (not legally necessarily but in terms of detection)?
Instead, I've been using this as an opportunity to teach people why big data is scary. In the past people often said they didn't care that Google could, and do, read their emails. Often responding with the ironic "I have nothing to hide" quote, or "I don't care, my stuff is boring." This pervasive rumor, of Facebook/Google always listening, has facilitated this conversation immensely. I've been able to better explain it to many who are not tech or statistic savvy in the least.
I don't believe Facebook is listening, because they don't need to.
And I believe all of them in fact would be perfectly fine with it - if it somehow helped them hit their quarterly bonuses or user growth numbers.
They could send just a transcript once every few minutes.
Another time I was on a road trip with a friend, driving through a food desert. We were talking about our food options -- stuff we never eat like McDonalds, Wendys, etc. Guess what pops up on Facebook moments later? An ad for Wendys! Were they listening? Or did they just connect the dots that it was lunch time, that we'd been driving for a long time without eating & that Wendys was one of the only options around?
In both instances, there were non-microphone explanations but creepy/impressive nonetheless.
I think I've got a lot of Facebook ads blocked, so I'm looking at a right hand column which is recommending me pages of random companies one of my friends has once liked and suggesting translations into two languages I don't speak which are spoken in countries I've never visited and have minimal connection with. Which I guess is an upgrade on all the dating ads which didn't even bother to match my "interested in"...
If you saw an ad for Wendy's in any other context, would you have been surprised?
I saw an ad for Wendy's shortly before coming to comment on this post. Are HN's advertising algorithms predicting the future, or is it just a coincidence due to the overwhelming volume of fast food advertising?
This is getting similar to horoscopes. When they are wrong, whatever. When they are right, it's "creepy".
designing an experiment to test this would be quite easy.
ultimately your explanation could exist alongside audio snooping, perhaps even helping to resolve areas of ambiguity...
That is - it is possible that the reason you were discussing the topic is the same reason that advertisers were targeting that particular phrase at that time. The Frequency Illusion could explain why you noticed that particular ad and why you were surprised. Also you presumably discuss topics all the time that you don't see ads for, but the probability of seeing an ad for any of those topics is quite high (the birthday problem).
[0] https://english.stackexchange.com/questions/153166/what-is-t...
It’s possible they might be doing it on Android, but once you show that a lot of people are imagining things, it’s a small leap to thinking that they all are.
The same way people believed iOS background apps couldn't record your screen? [0]
When you cannot trust the operating system manufacturer to be impartial, it stops being trivial to debunk.
[0] https://gizmodo.com/researchers-uber-s-ios-app-had-secret-pe...
It's all a little too much for me to believe it's a coincidence.
Even more troubling than audio recording in this case is that facebook has so much information about you that they don’t need it. The idea that facebook knows the approximate content of your conversations/thoughts to the point that it seems like they’re listening should be deeply troubling.
If facebook knows the content of your conversation, seemingly before or at the same time that you have it, what does that mean for free will? Agency? (Obviously free will is bs, but most people I’ve talked to refuse to acknowledge that - is this a wake-up call?)
What other predictions can they make? With the level of information given to facebook (browsing history, precise location, when you’re asleep and when you’re awake, and possibly much more, like your cell account info and the TV you watch) they can build a very complete picture of your life, and may even know things about yourself that you don’t know.
Of course, when facebook is actively listening to you (e.g. for facebook messenger) all bets are off. They already read your conversations anyway, so it’s not a far jump.
I think it’s hard to debunk because people don’t recognize how powerful their knowledge graph is at this point, and simultaneously underestimate how hard general-purpose speech recognition is, even with studio-level audio setups. And especially to a non-programmer, I think it seems so much easier to just figure out speech recognition than to maintain a huge knowledge graph about billions of people, their connections (explicit and implicit), their current whereabouts, the sites they are visiting, and the text they literally give to Facebook in the form of messages etc. But of course, each one of those things is a tractable problem, and can progress independently of the others, until all of a sudden it’s so good it seems like they must be inside our heads, or at least listening.
But here’s the thing: I think all that big data is actually waaaaay creepier. If they were just listening in and showing relevant ads, that’d be one thing. But instead they have all this data going back years, and can make all these inferences from the data. Just imagine what inferences could be made about non-advertising topics!
The people we talk with are generally the people we are connected with on social media. It's feasible to me that a conversation happens and someone in the group makes a search, which then maybe causes algorithms using location data to push those ads to people in that location.
Further, what's being talked about is probably being talked about and searched by others in the area and/or with the same interests, so even if it isn't localized quite so much the possibility of what's being talked about to generate ads is heightened.
Finally, there's the confirmation bias and recency bias at play. Once something is known or on mind, it's more likely to be noticed. They may have already had those ads coming in and only noticed them after talking about it. Or, the ads might have even caused the person to subconsiouciously think about the thing.
The psychology at play in advertising is advanced.
Source: https://www.forbes.com/sites/amitchowdhry/2017/10/31/faceboo...
The first one is that iOS gives users very fine grained control over what apps are allowed to use the microphone. It's not just a list of permissions, it asks specifically with a system dialogue box whether you want to allow an app to have permission.
Second, the status bar turns a very obvious red color when an app is using the microphone when it's in the background.
If FB was able to bypass both of those safeguards without Apple knowing, it would be a major story.
Think about it for a minute. Fb and others have so much data on people and have trained algos so well, even they don’t understand how they work so well. It’s not difficult to imagine this is the case; a computer, especially one with extensive data on the lifes, behaviors and thoughts of billions of people, is far smarter and better at predicting behavior than any human could.
Scary is that it’s just the beginning. I fall squarely into the alarmist camp on the AI issue. Not now, but in the long run. One can see how this scenario with the fb ad microphone spying paranoia thats actually 1 million IQ AI flexing its muscle is but the beginning of a slippery slope.
As long as there is no uncontestable proof, the doubt must benefit Facebook.
The testimonies are not uncontestbale proofs. Considering the million people uisng the facebook app, there will always be a small set of people who will see advertisement of what they just talked about. That is statistic. Some of these, understandably chocked, may be very vocal about it.
We also know that people are payed to spread fake news and manipulate people. So I'm cautious. The truth is unknown, and the longer it takes to get uncontestable proofs, the less credible the listening thesis becomes.
Facebook can't do much appart to let it rain. Open source is the way I would go. That is one way to calm down paranoid or conspirationist people.
they have a history of privacy violations
they have an incentive to snoop further
i see no reason to offer the benefit of the doubt to either hypothesis, as anecdotes do not constitute data
but for what it's worth, my hunch is that there's some extra snooping going on that has yet to be unearthed. might not be audio, necessarily.
Even so, i recon that your approach to test them is flawed, instead you should either of the following:
1. Reverse engineer the applications that has access to actually capture audio from the microphone to start with, i.e. the Messenger app on android.
2. Create a kernel module/microphone driver, which records when and which applications actually access the microphone. If the facebook apps are actually found accessing the microphone outside of a user initiated scope, one could bring the experiment to the next level and record the same audio recoding and save it for later review.
I recon 1. has already been done by several researchers, especially in relation to bug bounty programs etc. but that doesnt really mean that you should try to do it again for your own research.
Facebook and Google are using machine learning and massive data sets about the mental worlds we inhabit when we think nobody is looking.
And these algorithms are optimizing for clicks in so many vectors of psychological need and vulnerability that it’s already become incomprehensible for how they can know the things we haven’t yet understood we will think about.
They know about our conversations because they know the searches of the people we talk to, and what they’re thinking about. What will be on the top of their mind.
There’s nothing to go looking for or debunk. It’s just the next step in the data collection experiment.
I’m guessing they’ll need to tone down the effectiveness of ad targeting, if they haven’t already.
they use this information to employ a system of psychological levers which results in us willingly handing their proxies our money.
is it a good business practice. certainly. is it simple theft? my instinct is to say no, and instead ask:
is this new mastery of knowledge about peoples behaviors and the ability to affect other behaviors on a mass scale the most terrifying form of feudalism that mankind has known?
For me personally, the fact that they can learn so much about me is what's creepy and reason enough to not use their services. The how is merely an interesting bit of trivia.
A) Given Facebooks policy of A/B testing, it almost certainly isn't listening all the time.
and more importantly,
B) Now that everyone is talking about it, Facebook would have been stupid NOT to turn it off at least for now.
We've built AI that can do a great job of slicing and restitching your old conversation into responses that are plausible responses to input sentences.
Whether that's good at predicting a conversations depends a lot on the conversationalist you're attempting to emulate using that technique.
So why bother?
At the same time, it wouldn't take much at all for me to believe otherwise: a simple blinded test where some people leave their phones in a room, researchers come and talk about some subject (absent the owners), then leave without ever interacting with the owners of the phone.
There are two ways to do the next part: The owners of the phone 1 week later could be asked if they've seen any facebook ads on (topic). A control group would have to be compared (where researchers didn't speak on the topic.)
Alternatively, the subjects of the study could be given 5 random topics when commencing the study and asked to record any ads they've seen on any of the 5 topics. The researchers could then speak on 1 of the 5 topics (differing from person to person) and a statistical analysis could be performed.
A "proof" is a statistically valid correlation between what the researchers talked about out and what the blinded subjects reported seeing ads about.
A refutation is a lack of such a correlation. Easy, and it would convince me personally.
By the way based on the anecdotal evidence I strongly expect this study to conclude "facebook is listening."
-
Note: there are a few loose ends to take care of. The researchers who are tasked with speaking on a subject are more likely to google it on their own phones. (Having been exposed to the topic), and those are in geographic vicinity. There are other similar possible mechanisms.
Perhaps the best approach is if subjects' phones are in a sound-proof vault and the researchers' speech is either fed into it via speaker/microphone, or not done so, but the researchers do not know for any specific phone whether it is able to hear them. (Making the study "double-blind", as neither the subjects nor the researchers know whether the subjects' phones have heard anything on a subject.).
Without getting into the ads delivery part and the anecdotes, how would you prove that Facebook is listening? How would you prove that there is a set of information taken from your speech or your audible environment transferred to Facebook.
Also, your techniques could reveal a false negative both in the existence of a bug in the particular build of fb code being tested, and that such a mechanism could have existed in previous versions but already removed.
...goes on to describe multi-week, double-blind trials involving human subjects.
This is exactly why I believe pizzagate, am a 9/11 truther, etc.
There’s value in debunking a myth that seems plausible to the masses. In this case, the myth is plausible, and enough people without sufficient critical thinking skills believe it, therefore it’s worth debunking consistently at least within the tech community.
