A long time ago, when I was working on the initial implementation of android wear at Google, I actually worked with the code that does stuff like this, so I might be able to answer somewhat usefully. At least there, the way it worked, is that every 30 seconds, the device would wake up the accelerometer and collect 2 seconds of data at 120(?)Hz. After that, there was a relatively large decision tree based on the values, their derivatives, etc. This decision tree was an output of a large trained model, but was itself pretty small: a few thousand values. It could only classify things it was trained on - the output was an activity index. At Google at the time, the supported activities were: walking, biking, running, driving, sitting, unknown. The model cannot output anything other than the activity index.
The practical upshot: Could one detect such activities based on accelerometer data? Surely yes. However, unless somebody trained it on masturbation, it is unlikely that that is an actual possible output of it.
Details: model format was more or less this
node {
int activity; //positive if this is the terminal
//node and this is the answer, else this is
//not terminal. Then it is the index of the
//input sample to read (times minus one) to
//compare to the next value
float compareWith;
unsigned gotoNodeIdxIfLessThan;
unsigned gotoNodeIdxIfGreaterOrEq;
}
model {
node nodes[];
}
You’d start at node [0] and walk the tree as per comparison instructions (index of input samples and float to compare to) till you reached a terminal node.
Truthfully... Yes. Many other apps and services do as well. Careful what you give camera and mic permissions to, and always read the privacy agreements, especially for FAANG. If you don't think Amazon, Facebook, Apple, Google, and Microsoft are watching/listening (with and without the light indicators), think again.
I hear this refrain most often from non-technical people. My response is always that it's a waste of resources and unnecessary to listen to your conversations. Advertising companies can infer what you talk about with other people based on proximity to each other and subsequent actions like visiting a website.
If two people stand next to each other and talk about going to Las Vegas, and one person looks at flights to Last Vegas later on, you could probably show ads for flights to Las Vegas the person they were standing next to and have them click on it.
As someone who monitored services that participated in unprompted mic-popping and camera-popping, at-will, for a FAANG company, for multiple years, kindly shut the fuck up.
Additionally, don't be mad when you find out later that companies know about your hyper-personal tendencies.
Don't write off what you're hearing from your aunt as "FUD". What she's saying is a poor technical understanding of the very real overarching dynamic.
I've tried many times to explain to people things like that GPS satellites do not track you, and there are things you can do with your own phone which drastically cut down on tracking. There's no interest, no understanding that I'm simultaneously sharing their concerns while trying to tell them that they aren't powerless. They just think surveillance is a foregone conclusion, and then distance themselves and rationalize it any number of ways, and then share their nutty mechanism-of-action condemnations as a way of advertising their political tribe.
> "Don't write off what you're hearing from your aunt as "FUD". What she's saying is a poor technical understanding of the very real overarching dynamic."
I don't have poor technical understanding. I am pointing out one of the loopholes companies are using to exploit customers: "privacy agreements."
If you want an in-depth technical analysis of how big tech does spy on you, just ask. I can provide a few case studies using OSINT and "hypothetical" speak, so as not to break NDA.
Sorry. I wasn't saying that your comment was based on poor technical understanding. I was making that judgement about GP's strawman of your comment.
I'd be interested in hearing what you have to say from the technical side, in case it contains any information that is new to me. Personally, I model things such that if a closed software/system has access to some information, it will backhaul, store, and abuse that information (if not immediately, then in the future). I'm not going to hang my hat on a hope that some companies aren't functioning as optimal surveillance machines!
FWIW I don't find privacy policies themselves particularly enlightening because they're generally vague, equivocating, and subject to change at whim. Sure technically if there weren't a "privacy policy" then you might theoretically have some legal right of action to go after a company for abusing your information (although good luck demonstrating real damages, as always). But at least in the US that feels completely hypothetical.
Do you find it weird/alarming that you assert a female assignment and a status of "poor technical understanding"? Where do you work? What is your role? Are you in a leadership position? Asking on behalf of everyone who does not want to work near your flagrant bias.
Tell me you have a way of proving that they're not listening to you all the time through your phone. I don't, the average person certainly doesn't. But it's certainly physically possible, not like I can prove these permissions controls do what they say in my phone, and every couples of months out comes some news about how you car was tracking you all these years.
I can tell you that I have a pretty good idea that the application software and operating system on my phone isn't listening to me all the time. I have a pretty good idea that the permissions controls (capabilities) do what they say. This is because I run libre Graphene OS on my phone, rather than a proprietary Android.
(I cannot prove that the baseband chipset/software is betraying me by listening all the time. The point is you have to dig through most of the layers of avoidable surveillance to focus on the real technical flaws)
> every couples of months out comes some news about how you car was tracking you all these years.
Please try not to take this too personally, but this is the effect I'm talking about where it feels like all this complexity gets bundled into learned helplessness. Does your car have a cell modem? It so, "it is" tracking your location - meaning the cell provider(s) are obtaining your location, and most likely selling it to the surveillance industry, in addition to whatever your car manufacturer is doing. If you want to be assured that it's not, remove the cell modem or buy an older car without built in surveillance.
(I'm certainly not putting forth self-help options as a substitute for needed overarching privacy legislation like the EU's GDPR. They're both independent topics worthy of discussion)
"Very long time Echo user starting to have a pretty negative (and questionable) experience"
"Over the past couple months, I’ve noticed all of our more recently purchased devices have suddenly started reacting very erratic with voice commands. It’s not one or two devices in particular, but almost all of them. Saying ‘Alexa’ even when standing next to the Show 10 will oftentimes trigger responses to a device across the house as if the spacial detection is off, but it makes no sense as the device reacting couldn’t have possibly heard you. Additionally, we’ve had a lot of false activations and pulling up the history usually presents “Audio was not intended for this device” or “audio could not be understood”, but the recordings have absolutely nothing to do with Alexa, nor is anything remotely close to it present, and sometimes there simply isn’t anything listed for review. In the past the false triggers were uncommon, but when they occurred, they were consistently presented in the app to review and generally made sense. I even had a situation this week where the device in my kitchen kept activating and I was repeatedly yelling “cancel” and the damn thing kept going. I ended up muting the mic for a day as it was so frustrating."
The practical upshot: Could one detect such activities based on accelerometer data? Surely yes. However, unless somebody trained it on masturbation, it is unlikely that that is an actual possible output of it.
Details: model format was more or less this
You’d start at node [0] and walk the tree as per comparison instructions (index of input samples and float to compare to) till you reached a terminal node.