“What would it mean for your business if you could target potential clients who are actively discussing their need for your services in their day-to-day conversations? No, it's not a Black Mirror episode—it's Voice Data, and CMG has the capabilities to use it to your business advantage.”
I usually wear the tin foil hat in these debates, but I must concede in this case: the eavesdropping phone theory in particular is difficult to substantiate, from a technical standpoint.
For one, a user can check this themselves today with basic local network traffic monitors or packet sniffing tools. Even heavily compressed audio data will stand out in the log, no matter how it’s encrypted, streamed, batched or what have you.
To get a sense of what I mean, run wireshark and give a wake phrase command to see what that looks like. Now imagine trying to obfuscate that type of transmission for audio longer than 2 seconds, and repeatedly throughout a day.
Even assuming local audio inference and processing on a completely compromised device (rooted/jailbroken, disabled sandboxing/SIP, unrestricted platform access, the works) most phones will just struggle to do that recording and processing indeterminately without a noticeable impact on energy and data use.
I’m sure advertising companies would love to collect that much raw candid data. It would seem quite a challenge to do so quietly, however, and given the apparent lack of evidence, is thus unlikely to have been implemented at any kind of scale.
There’s also a totally plausible and far more insidious answer to what’s going on with the experiences people have of the ads matching their conversations.
That explanation is advertising works. And worse, it works subconsciously. That you’re seeing the ads and don’t even notice you’re seeing them and then they’re worming their way into your conversations at which point you become more aware of them and then start noticing the ads.
Which does comport with the billions of dollars spent on advertising every year. It would be very weird if an entire ad industry that’s at least a century old was all a complete nonsense waste of money this whole time.
To me, this whole narrative is just another parable about why we need to do everything possible to limit our own exposure to ads to avoid being manipulated.
Damn, I hadn’t thought of that. The chicken egg question of spooky ad relevance. Insidious indeed.
I feel like the idea of some person or group having enough info to psychologically manipulate or predict should be way scarier than the black helicopter stuff, especially given that it’s one of the few conspiracy theories we actually have a bunch of high quality evidence for, just in marketing and statistics textbooks alone.
But here we are. Government surveillance is the hot button, not the fact that marketers would happily sock puppet you given the chance.
That is glossing over how they process the data and transmit it to the cloud. The assistant wake word for “Hey Google” invokes an audio stream to an off site audio processor in order to handle the query. So that is easy to identify via traffic because it is immediate and large.
The advertising-wake words do not get processed that way. They are limited in scope and are handled by the low power hardware audio processor used for listening for the assistant wake word. The wake word processor is an FPGA or ASIC - specifically because it allows the integration of customizable words to listen for in an extremely low power raw form. When an advertising wake word is identified, it sends an interrupt to the CPU along with an enumerated value of which word was heard. The OS then stores that value and transmits a batch of them to a server at a later time. An entire day’s worth of advertising wake word data may be less than 1 kb in size and it is sent along with other information.
Hmm, that’s outside my wheelhouse. So you’re saying phone hardware is designed to listen for not just one but multiple predefined or reprogrammable bank of wake words? I hadn’t read about that yet but it sounds more feasible than the constant livestream idea.
The echo had the capacity for multiple wake words IIRC, but I hadn’t heard of that for mobile devices. I’m curious how many of these key words can they fit?
Smartphones by definition are Spyware, at least if you use the OS as is, because in them all aspects are controlled and logged, either by Google on Android or by Apple on iOS. Adding the default apps that cannot be uninstalled on a mobile that is not rooted.
As COX alleges, they also use third-party logs and therefore can track and profile the user very well, even without using this technology that they claim to have.
Although they feel authorized by the user’s consent to the TOS and PP, the legality depends directly on the legislation of each country. TOS and PP itself, to be a legal contract, must comply in all its points with local legislation to be applicable to the user. For this reason, I think that these practices are very different in the EU from those in the US, where legislation regarding privacy is conspicuous by its absence, that is, that US users should take these COX statements very seriously in their devices, although in the EU they must also be clear that Google and Apple know exactly what they do and where users live, although they are limited from selling this data to third parties.
Basics:
– READ ALWAYS TOS AND PP
Review the permissions of each app, leaving only the most essential ones
Desactivate GPS if not used
Review in Android every app with Exodus Privacy, maybe Lookout or MyCyberHome in iOS (Freemium apps !!!)
Use as less possible apps from the store
Be aware of discount apps from the Supermarket or Malls
Don’t store important data in the Phone (Banking, Medical…)
Agreed, though I think it’s possible to use smart devices safely. For Android it can be difficult outside custom roms. The OEM flavors tend to have spyware baked in that takes time and root to fully undo, and even then I’m never sure I got it all. These are the most common phones, however, especially in economy price brackets, which is why I’d agree that for the average user most phones are spyware.
Flashing is not useful advice to most. “Just root it bro” doesn’t help your nontechnical relatives who can’t stop downloading toolbars and VPN installers. But with OEM variants undermining privacy at the system level, it feels like a losing battle.
I’d give credit to Apple for their privacy enablement, especially with E2EE, device lockdown, granular access permission control and audits. Unfortunately their devices are not as affordable and I’m not sure how to advise the average Android user beyond general opt-out vigilance.
Yeah those push token systems need an overhaul. IIRC tokens are specific to app-device combinations, so invalidation that isn’t automatic should be push-button revocation. Users should have control of it like any other API on their device, if only to get apps to stop spamming coupons or whatever.
It’s funny though: when I first saw those headlines, my first reaction was that it was a positive sign, since this was apparently news worthy even though the magnitude of impact for this sort of systemic breach is demonstrably low. (In particular, it pertains to (1) incidental high-noise data (2) associated with devices and (3) available only by request to (4) governments, who are weak compared to even the smallest data brokers WRT capacity for data mining inference and redistribution, to put it mildly.)
My own theory is that they tokenize key words and phrases with an AI so that they’re not sending the actual audio data. Then it’s stored in a form some AI can parse but isn’t technically user data so they can skirt legislation around that.
A tokenized collection of key phrases omitting delimiters in text format is going be much, much less than audio, or a transcript.
as someone who has played around with offline speech recognition before - there is a reason why ai assistants only use it for the wake word, and the rest is processed in the cloud: it sucks. it’s quite unreliable, you’d have to pronounce things exactly as expected. so you need to “train” it for different accents and ways to pronounce something if you want to capture it properly, so the info they could siphon this way is imho limited to a couple thousand words. which is considerable already, and would allow for proper profiling, but couldn’t capture your interest in something more specific like a mazda 323f.
but offline speech recognition also requires a fair amount of compute power. at least on our phones, it would inevitably drain the battery
What if its not streaming? What if its just cached for future access, e.g. next time the user opens the app (and network traffic spikes anyways) maybe?
That’s possible too, and in general I’d think a foreground application currently in use alleviates most of the technical restrictions mentioned (read: why we never install FB).
But again we must assume some uncommon device privileges and we still haven’t solved the problem of background energy usage required to record and/or process a real time feed.
Can it be implemented on pc? They often turned on and people speak around them too. Cpu activity much harder to trace when there are a lot of different processes.
Someone can blame their phone, while it listening pc near by.
Yeah outside mobile devices I imagine there’s a lot more leeway technically speaking. I’d be far more inclined to suspect a smart TV or a home assistant appliance like Amazon Echo, for example. And certainly there are plenty of PCs out there that are 100% compromised.
But it’s the phone that people often think of as eavesdropping on their conversations. The idea is stickier perhaps because it’s a more personal violation. And I wouldn’t put it past data brokers by any means. They would if they could. I’ve just yet to hear a feasible explanation of how they can without being caught. Hence my doubt.
most phones will just struggle to record and process audio indeterminately without a noticeable impact on energy and data use.
I mean, it’s still a valid concern for a commoner. Why my phone has twice the ram and twice the cores and is as slow as my previous one? I’d love to fuel this conspiracy into OS, app makers to do their fucking job.
There’s no reason an app can weight more than 50mb on clean install*, and many socials, messengers fail to fit in. A client I use to write this is only 30+, and that’s one person doing that for donations.
If there could be a raging theory that apps are selling your data to, like, China, there would be a push to decline it and optimize apps to fit that image.
* I obviously exclude games, synths, editors of any kind with their textures and templates.
The filesize of most binaries is dominated by text strings and images. Modern applications are loaded with them. Lemmy is atypical in that it doesn’t need tons of built in images or text.
I get it. It’s just I don’t see any dev-put images in many big apps, besides a logo and a welcome screen. Updating them with dozens of megabytes doesn’t feel okay. It seems like there’s some bloat, or a vault management problems. Like in some seasonally updated games that put dupes to speed up load of a map or easily add new content on top of them instead of redownloading a brand new db. Some I followed shawed off tens of gigabytes by rearranging stuff.
Like, messengers. I don’t get it how Viber wants more than 40+ mb per update having nothing but stickers, emoji already installed and probably don’t change them much. Cheap wireless connection could allow them to ignore that for some reason and start to get heavier in order to offload some from their servers, for many images are localized. Is that probably what their updates are? Or they consequentially add beta patches after an approval, so you download a couple of them in a close succession after they get into public?
Yeah they’d have to it seems, but real time transcription isn’t free. Even late model devices with better inference hardware have limited battery and energy monitoring. I imagine it’d be hard to conceal that behavior especially for an app recording in the background.
[email protected] mentioned that mobile devices use the same hardware coprocessing used for wake word behavior to target specific key phrases. I don’t know anything about that, but it’s one way they could work around the technical limitations.
Of course, that’s a relatively bespoke hardware solution that might also be difficult to fully conceal, and it would come with its own limitations. Like in that case, there’s a preset list of high value key words that you can tally, in order to send company servers a small “score card” rather than a heavy audio clip. But the data would be far less rich than what people usually think of with these flashy headlines (your private conversations, your bowel movements, your penchant for musical theater, whatever).
What could be possible, would be maybe send tiny bits. For example, a device could categorize some places or times, detect out of pattern behaviours and just record a couple of seconds here and there, then send it to the server when requesting something else to avoid being suspicious. Or just pretend it’s a “false positive” or whatever and say “sorry, I didn’t get that.”
I don’t think they’re listening to everything, but they could technically get something if they wanted to target you.
Right, I suppose cybersecurity isn’t so different than physical security in that way. Someone who really wants to get to you always can (read: why there are so many burner phones at def con).
But for the average person, who uses consumer grade deadbolts in their home and doesn’t hire a private detail when they travel, does an iPhone fit within their acceptable risk threshold? Probably.
I usually wear the tin foil hat in these debates, but I must concede in this case: the eavesdropping phone theory in particular is difficult to substantiate, from a technical standpoint.
For one, a user can check this themselves today with basic local network traffic monitors or packet sniffing tools. Even heavily compressed audio data will stand out in the log, no matter how it’s encrypted, streamed, batched or what have you.
To get a sense of what I mean, run wireshark and give a wake phrase command to see what that looks like. Now imagine trying to obfuscate that type of transmission for audio longer than 2 seconds, and repeatedly throughout a day.
Even assuming local audio inference and processing on a completely compromised device (rooted/jailbroken, disabled sandboxing/SIP, unrestricted platform access, the works) most phones will just struggle to do that recording and processing indeterminately without a noticeable impact on energy and data use.
I’m sure advertising companies would love to collect that much raw candid data. It would seem quite a challenge to do so quietly, however, and given the apparent lack of evidence, is thus unlikely to have been implemented at any kind of scale.
There’s also a totally plausible and far more insidious answer to what’s going on with the experiences people have of the ads matching their conversations.
That explanation is advertising works. And worse, it works subconsciously. That you’re seeing the ads and don’t even notice you’re seeing them and then they’re worming their way into your conversations at which point you become more aware of them and then start noticing the ads.
Which does comport with the billions of dollars spent on advertising every year. It would be very weird if an entire ad industry that’s at least a century old was all a complete nonsense waste of money this whole time.
To me, this whole narrative is just another parable about why we need to do everything possible to limit our own exposure to ads to avoid being manipulated.
Damn, I hadn’t thought of that. The chicken egg question of spooky ad relevance. Insidious indeed.
I feel like the idea of some person or group having enough info to psychologically manipulate or predict should be way scarier than the black helicopter stuff, especially given that it’s one of the few conspiracy theories we actually have a bunch of high quality evidence for, just in marketing and statistics textbooks alone.
But here we are. Government surveillance is the hot button, not the fact that marketers would happily sock puppet you given the chance.
That is glossing over how they process the data and transmit it to the cloud. The assistant wake word for “Hey Google” invokes an audio stream to an off site audio processor in order to handle the query. So that is easy to identify via traffic because it is immediate and large.
The advertising-wake words do not get processed that way. They are limited in scope and are handled by the low power hardware audio processor used for listening for the assistant wake word. The wake word processor is an FPGA or ASIC - specifically because it allows the integration of customizable words to listen for in an extremely low power raw form. When an advertising wake word is identified, it sends an interrupt to the CPU along with an enumerated value of which word was heard. The OS then stores that value and transmits a batch of them to a server at a later time. An entire day’s worth of advertising wake word data may be less than 1 kb in size and it is sent along with other information.
Good luck finding that on wireshark.
Hmm, that’s outside my wheelhouse. So you’re saying phone hardware is designed to listen for not just one but multiple predefined or reprogrammable bank of wake words? I hadn’t read about that yet but it sounds more feasible than the constant livestream idea.
The echo had the capacity for multiple wake words IIRC, but I hadn’t heard of that for mobile devices. I’m curious how many of these key words can they fit?
deleted by creator
Smartphones by definition are Spyware, at least if you use the OS as is, because in them all aspects are controlled and logged, either by Google on Android or by Apple on iOS. Adding the default apps that cannot be uninstalled on a mobile that is not rooted. As COX alleges, they also use third-party logs and therefore can track and profile the user very well, even without using this technology that they claim to have.
Although they feel authorized by the user’s consent to the TOS and PP, the legality depends directly on the legislation of each country. TOS and PP itself, to be a legal contract, must comply in all its points with local legislation to be applicable to the user. For this reason, I think that these practices are very different in the EU from those in the US, where legislation regarding privacy is conspicuous by its absence, that is, that US users should take these COX statements very seriously in their devices, although in the EU they must also be clear that Google and Apple know exactly what they do and where users live, although they are limited from selling this data to third parties.
Basics:
– READ ALWAYS TOS AND PP
Agreed, though I think it’s possible to use smart devices safely. For Android it can be difficult outside custom roms. The OEM flavors tend to have spyware baked in that takes time and root to fully undo, and even then I’m never sure I got it all. These are the most common phones, however, especially in economy price brackets, which is why I’d agree that for the average user most phones are spyware.
Flashing is not useful advice to most. “Just root it bro” doesn’t help your nontechnical relatives who can’t stop downloading toolbars and VPN installers. But with OEM variants undermining privacy at the system level, it feels like a losing battle.
I’d give credit to Apple for their privacy enablement, especially with E2EE, device lockdown, granular access permission control and audits. Unfortunately their devices are not as affordable and I’m not sure how to advise the average Android user beyond general opt-out vigilance.
https://arstechnica.com/tech-policy/2023/12/apple-admits-to-secretly-giving-governments-push-notification-data/
Yeah those push token systems need an overhaul. IIRC tokens are specific to app-device combinations, so invalidation that isn’t automatic should be push-button revocation. Users should have control of it like any other API on their device, if only to get apps to stop spamming coupons or whatever.
It’s funny though: when I first saw those headlines, my first reaction was that it was a positive sign, since this was apparently news worthy even though the magnitude of impact for this sort of systemic breach is demonstrably low. (In particular, it pertains to (1) incidental high-noise data (2) associated with devices and (3) available only by request to (4) governments, who are weak compared to even the smallest data brokers WRT capacity for data mining inference and redistribution, to put it mildly.)
Regardless, those systems need attention.
My own theory is that they tokenize key words and phrases with an AI so that they’re not sending the actual audio data. Then it’s stored in a form some AI can parse but isn’t technically user data so they can skirt legislation around that.
A tokenized collection of key phrases omitting delimiters in text format is going be much, much less than audio, or a transcript.
as someone who has played around with offline speech recognition before - there is a reason why ai assistants only use it for the wake word, and the rest is processed in the cloud: it sucks. it’s quite unreliable, you’d have to pronounce things exactly as expected. so you need to “train” it for different accents and ways to pronounce something if you want to capture it properly, so the info they could siphon this way is imho limited to a couple thousand words. which is considerable already, and would allow for proper profiling, but couldn’t capture your interest in something more specific like a mazda 323f.
but offline speech recognition also requires a fair amount of compute power. at least on our phones, it would inevitably drain the battery
That certainly would make the data smuggling easier. What about battery though? I assume that requires inference and at least rudimentary processing.
How would a background process do this in real time on a mobile device without leaving traceable evidence like cpu time?
What if its not streaming? What if its just cached for future access, e.g. next time the user opens the app (and network traffic spikes anyways) maybe?
Or plugs in their phone at night, bypassing energy use concerns?
That’s possible too, and in general I’d think a foreground application currently in use alleviates most of the technical restrictions mentioned (read: why we never install FB).
But again we must assume some uncommon device privileges and we still haven’t solved the problem of background energy usage required to record and/or process a real time feed.
Cox also sells home automation bundles which advertise “smart” features like voice recognition which are always plugged into the wall.
deleted by creator
Can it be implemented on pc? They often turned on and people speak around them too. Cpu activity much harder to trace when there are a lot of different processes. Someone can blame their phone, while it listening pc near by.
Yeah outside mobile devices I imagine there’s a lot more leeway technically speaking. I’d be far more inclined to suspect a smart TV or a home assistant appliance like Amazon Echo, for example. And certainly there are plenty of PCs out there that are 100% compromised.
But it’s the phone that people often think of as eavesdropping on their conversations. The idea is stickier perhaps because it’s a more personal violation. And I wouldn’t put it past data brokers by any means. They would if they could. I’ve just yet to hear a feasible explanation of how they can without being caught. Hence my doubt.
I mean, it’s still a valid concern for a commoner. Why my phone has twice the ram and twice the cores and is as slow as my previous one? I’d love to fuel this conspiracy into OS, app makers to do their fucking job.
There’s no reason an app can weight more than 50mb on clean install*, and many socials, messengers fail to fit in. A client I use to write this is only 30+, and that’s one person doing that for donations.
If there could be a raging theory that apps are selling your data to, like, China, there would be a push to decline it and optimize apps to fit that image.
* I obviously exclude games, synths, editors of any kind with their textures and templates.
The filesize of most binaries is dominated by text strings and images. Modern applications are loaded with them. Lemmy is atypical in that it doesn’t need tons of built in images or text.
I get it. It’s just I don’t see any dev-put images in many big apps, besides a logo and a welcome screen. Updating them with dozens of megabytes doesn’t feel okay. It seems like there’s some bloat, or a vault management problems. Like in some seasonally updated games that put dupes to speed up load of a map or easily add new content on top of them instead of redownloading a brand new db. Some I followed shawed off tens of gigabytes by rearranging stuff.
Like, messengers. I don’t get it how Viber wants more than 40+ mb per update having nothing but stickers, emoji already installed and probably don’t change them much. Cheap wireless connection could allow them to ignore that for some reason and start to get heavier in order to offload some from their servers, for many images are localized. Is that probably what their updates are? Or they consequentially add beta patches after an approval, so you download a couple of them in a close succession after they get into public?
What if the processing is done locally and the only thing they send back home is keywords for marketable products?
Yeah they’d have to it seems, but real time transcription isn’t free. Even late model devices with better inference hardware have limited battery and energy monitoring. I imagine it’d be hard to conceal that behavior especially for an app recording in the background.
[email protected] mentioned that mobile devices use the same hardware coprocessing used for wake word behavior to target specific key phrases. I don’t know anything about that, but it’s one way they could work around the technical limitations.
Of course, that’s a relatively bespoke hardware solution that might also be difficult to fully conceal, and it would come with its own limitations. Like in that case, there’s a preset list of high value key words that you can tally, in order to send company servers a small “score card” rather than a heavy audio clip. But the data would be far less rich than what people usually think of with these flashy headlines (your private conversations, your bowel movements, your penchant for musical theater, whatever).
deleted by creator
Fucking thank you. As I said in another reply, if this was true my firewall logs would be full, or my data cap blown in a week.
I agree.
What could be possible, would be maybe send tiny bits. For example, a device could categorize some places or times, detect out of pattern behaviours and just record a couple of seconds here and there, then send it to the server when requesting something else to avoid being suspicious. Or just pretend it’s a “false positive” or whatever and say “sorry, I didn’t get that.”
I don’t think they’re listening to everything, but they could technically get something if they wanted to target you.
Right, I suppose cybersecurity isn’t so different than physical security in that way. Someone who really wants to get to you always can (read: why there are so many burner phones at def con).
But for the average person, who uses consumer grade deadbolts in their home and doesn’t hire a private detail when they travel, does an iPhone fit within their acceptable risk threshold? Probably.