- There is not evidence that these devices record and transmit without an activation word triggering this behavior
- However, there is nothing to stop companies from breaking this assumption
- Some people think the risk of one of these companies flipping a switch and recording everything is negligible
- Some people think the risk of one of these companies flipping a switch and recording everything is warrants serious concern
- These two groups will not agree, and that's fine :)
There is hard evidence [1][2] you can remotely operate Echo recording capabilities without a wake word. Hope this puts the 'hardware limitation' claim to bed.
It's like in school where there's an advantage/disadvantage question and I only know one thing: I wore the same thing as advantage and disadvantage.
The advantage of a stationary device is you don't have to charge it. It is always connected.
The disadvantage is you don't have to charge it. It is always connected so the engineers don't have to make trade offs they'd have to make on a battery operated phone.
Nice summary. For those who believe that one of these companies might (intentionally or accidentally) "flip the switch", would a project like this really do that much to persuade you that the device had now become safe for use? Or would you simply avoid knowingly purchasing any such devices? (In that sense I'm struggling to understand the true customer for a neat hack like this.)
While I appreciate the sentiment...unless you actually think Google and Amazon devices are recording irrelevant ambient sound deliberately (they aren’t), this doesn’t help anything. Unless the software here is better than theirs at recognizing the trigger word (very unlikely), there will be even more false positive activations on this device than there are on the originals.
Edit: It’s very unlikely because Amazon and Google pay for false positives, so they have a strong incentive to develop really good trigger word detection.
How do you know? And, how do you know they will not do this silently in the future?
Also worse detection does not mean more false positives. Usually, you can get the false positive rate very low by allowing more false negatives. In this way you have a choice, how you want to trade-off. Without this device, you are stuck with the choice that Amazon/Google make for you.
> How do you know? And, how do you know they will not do this silently in the future?
Because it's a literal hardware limitation. The device is built in a way that requires a wake word before any recording can possibly happen, thanks to it being built with 2 separate control boards. If they ended up maybe changing the wakeword to "the", then maybe they could "silently" listen to everything, but that would be caught pretty quick because the device would be "lit up" constantly (another _hardware_ thing), or someone would notice that it no longer responds to "Alexa" or "Google".
Seriously, a lot of people on HN need to do their damn homework about these devices before declaring them to be something they have been proven not to be. Packet sniffing and hardware inspection both instantly disprove all these conspiracy-theory nonsense claims that these devices are recording your every word.
Those two separate control boards didn't stop my Amazon dot from acually recording ambient noise and uploading it to Amazon's systems. I know this because of the audio history they themselves provide! You can literally go back and play back all the audio recorded, and a great deal of it did not include questions. Further, there was also a report of being able to trigger audio recording without either activating the LED ring or using a wake word via a serial root console. While a third-party attacker is unlikely to use that method of access, nothing about the hardware actively prevents Amazon ftom triggering it that way. Likewise for Google.
And yes, I work on this stuff. Neither Google nor Amazon have the hardware limitations you suggest.
You are making an enormous amount of assumptions based on a semantic argument.
Echo devices only begin recording if they think they hear the wake word. Obviously this is less than straight-forward, hence the recordings that didn't follow the wake word (just examples of an Alexa device incorrectly thinking it heard it).
To suggest that a serial root console is a point of attack for an Echo device is bordering on insanity. You'd need a breakout board connected via the USB interface (not port, mind you) in order for this work-around to be effective. So yes, if a hacker had physical access to your device, time enough to solder on a breakout board, said third-party could record a variety of things.
But then, it's a whole hell of a lot easier to just install a mic in someones house and get the same effect, now wouldn't it?
> To suggest that a serial root console is a point of attack for an Echo device is bordering on insanity.
That was not what he said. He argues that Amazon/Google could remotely use a similar exploit (without direct access to the hardware) to start recording without lighting up the LED.
Nobody has EVER gotten root console access on an Echo device remotely, and the only successful "remote" exploit that didn't require soldering requires that the attacker and the victim are both on the same wifi network.
Please, feel free to explain how Amazon and Google could exploit that vulnerability (that has since been patched)? More importantly, I'd love to hear how they are going to pull this off and hide it, given network traffic will be a dead give away?
If what your suggesting is actually what he meant, that's even more absurd than attackers trying to do the same.
True enough. They could easily push a new update that would record every single thing you say, and despite not indicating anywhere on the device, it would take a matter of minutes before it was in the news because what they certainly can't do is hide network traffic.
As indicated in your previous comments, e.g. https://news.ycombinator.com/item?id=18616219 , you work for Amazon. It would be a better look if you disclosed this openly when commenting about Amazon.
It's easily located in my post history, however, I don't work with anything even remotely related to the Echo devices. My interest in this discussion is as a user, not as an employee.
It's an ethics issue. You have vested interest in Amazon's public perception as an employee. You can't try to divorce your comments from your relationship with Amazon and expect to still be taken seriously.
A simple disclosure would have lent your comments more credibility.
> Because it's a literal hardware limitation. The device is built in a way that requires a wake word before any recording can possibly happen [...]
Take note that Amazon Drop In [1] is a feature built around turning on the Echo mic remotely without a wake word. I don't think this feature could exist if there was a hardware limitation.
I was just offering a counter to the hardware limitation claim. It's possible the device makes itself known during use, I haven't used this feature yet.
These are software companies. These devices support OTA updates, including changing the activation word. It's trivial to change the activation word to empty or something innocuous. Ergo, there's no hardware limitation.
If your argument stems from "Google and Amazon would never do that," I do not trust any corporate entity to value my rights more than their ability to make a dollar.
How about the issue where Google’s devices were errantly recording everything due to a hardware issue - where the button override for the voice activation was stuck in the activated position.
People who think that there’s no way that Google and Amazon could be recording everything need to realize that this is also not true. Most of these “limitations” are software enforced, and that software is updated constantly.
This is the main problem that I see. Sure, I tested the packets, sniffed them, made sure it wasn't recording, etc, but then they push an update the next day. I don't think it's practical to monitor these devices all the time, and I haven't been asked to opt-in to an Echo update.
I also don't necessarily assume mal-intent on the part of the companies, but that doesn't mean there won't _ever_ be that intent. Trusting that all of these assumptions hold over time is hard.
"Malintent" can be a hard bar to clear, but it's clear beyond a shadow of a doubt that these companies view these devices as mechanisms to push forward their own interests and desires, in addition to my own. I won't even necessarily call that morally wrong, or at least, that line is very fuzzy. But it does mean that viewing them with a certain amount of suspicion is just rational, not crazytalk.
(It's true of cell phones too, of course, and I am engaged in constant activity to ensure the phone works for me, and not any of the many corporations that want to make it work for them. Turning off notifications, uninstalling certain apps after they've gone bad, ensuring permissions aren't too wide open, uninstalling default-installed apps and disabling others... it's a constant battle made worthwhile only by the fact that in the end, I really have mostly mastered my phone and it is working for me. I don't have one of these audio assistants because it is far less clear to me how to do that. Modulo being spied on by intelligence agencies, anyhow, although at this point I'm not sure how one could even escape that.)
> This is the main problem that I see. Sure, I tested the packets, sniffed them, made sure it wasn't recording, etc, but then they push an update the next day. I don't think it's practical to monitor these devices all the time, and I haven't been asked to opt-in to an Echo update.
This is a problem with forced updates in general (I'm also thinking of Windows, Chrome, Chrome extensions, etc. here) that security experts seem completely blind to.
That said, note that even if the software didn't update, it doesn't mean it would have to send bad packets when you're actually observing. It could randomly start doing that once in a while after a few months.
> Trusting that all of these assumptions hold over time is hard.
This is the big point. You are not only trusting that the company as it is today is doing the right thing, but that the company will continue to do the right thing for so long as the device is in your house - and that they will do the right thing in perpetuity with your data (including if/when they sell the company down the road).
I'm not sure how you would regress a button that physically doesn't exist anymore. Also, if you're that scared of future bugs that don't exist, then you should probably throw away your smart phone.
I think the point is that bugs exist and will continue to exist: whether it's the same bug, a different one, mal-intent, negligence, or anything else. Sure, this one device won't solve every single problem out there, but should we not solve anything just because we can't solve everything?
Right, and my point is that bugs will exist for all devices, not just these. Applying the logic, "bugs could happen" to just these devices isn't rational because it applies to all all devices, especially smart phones. We shouldn't ditch devices because of future potential for bugs that don't exist yet.
The entire top surface of the Google Home is a button (capacitive). Those kinds of sensors are just as susceptible to physical defects as mechanical buttons.
As a side note, whataboutism adds nothing of value to this discussion about the Google Home and Amazon Echo.
And the capacitive button doesn't trigger the listening hardware. Splitting hairs over the hardware specifications isn't proof that the bug is still a problem.
Also, talking about your contradictory behavior with your smartphone isn't whataboutism, unless you want to avoid addressing your hypocrisy, because smart phones are susceptible to the same blanket fears you have with homes/alexa. To critique only the latter, and not the former (which you use daily), is not fair.
"The device is built in a way that requires a wake word before any recording can possibly happen, thanks to it being built with 2 separate control boards."
This is a very technically naive interpretation of their hardware/software solution.
If the wake word were hard-coded into silicon then perhaps I would be charitable about your misunderstanding(s) - but of course it is not. The wake word is user-definable and can be changed to arbitrary sounds at any time.
Whatever hardware limitation(s) may exist are trivially worked around with software, which can be updated over the top of you at any time.
There's a widespread sentiment that current evidence of compliance to "doing right by users" should be viewed with circumspection. And it's fair to say Google's past behaviour raises doubts about the level of trust users should extend them.
What is a conspiracy theory about today's hardware, I have no trouble imagining is a planned or at least considered future iteration of their "service".
> Seriously, a lot of people on HN need to do their damn homework about these devices before declaring them to be something they have been proven not to be.
Based on what? Marketing copy? Eyeballing iFixit teardowns?
> but that would be caught pretty quick because the device would be "lit up" constantly (another _hardware_ thing)
Totally not buying it, unless you can show me the traces and discrete components that force power through the LED when signal from the microphone is allowed to reach the uC. If it's done in software, I'm not trusting it.
If you're making an argument of "trust the vendor because economics", you have to recognize how weak it is.
Do you (or anyone else with the same claim) have a citation for this?
I've spent some time reverse-engineering the echo microphone board, and while there is an interlock that prevents recording while the red mute button is lit (just the light under the button, not the ring), I didn't see anything that would prevent recording while the ring light was off.
Unless there is a separate out of band board with a relay I can hear or see (meaning, code alone can't enable something), then it really isn't a hardware limitation. The security controls and operations are in the code. The code can change or may already have silent monitoring capabilities. Nobody on HN could really answer whether or not this is the case. All we can do is speculate. If someone were required to put lawful monitoring code in place, they would not be allowed to discuss it here. The best anyone could do is decompile the code or get the source code for the firmware. Even then, there could be non-volital space that allows for updates.
Case in point, there have been malware packages that could enable your microphone and camera on the laptop without turning on the LED. This varied with camera model. Some power the LED when the camera has power. Microphones don't always activate an LED. There are a myriad of articles you can find providing examples of malware that can listen to cell phone microphones, laptop microphones without activating the LED.
How so? Ultimately the mic is always on and listening for its keywords - if you look at the teardown of the Alexa on iFixIt, I don't even see any device other than the main CPU that would be capable of performing keyword recognition. Meaning the main CPU would have to be the thing then controlling the lights after the keywords are recognized...
The Google Home at least has a separate board with a microcontroller on it which could be used for keyword recognition, but I'm pretty sure they allow that to be updated for the sake of improving keyword recognition and there's no reason that an update couldn't disable the LEDs in the listen state as far as I can see.
Yes, I don't mean to say it's impossible - just that you'd need an entirely isolated system to detect when data was flowing over that link which is physically connected in all cases and cannot be updated. I don't believe we see that in either the Alexa or Google Home, but I'd be happy to be mistaken if anyone's done a more in depth teardown of these systems.
And all of this is hinged on hoping you notice LEDs firing in the corner while you're having a conversation. Perhaps a more noticeable method should be used in cases like this. A forced "beep"/tone or something from an isolated circuit hardwired to the speakers.
As an alternate angle - Instead of trying to disable the light have it show the "I'm doing a software update" light pattern. I know I personally wouldn't give that a second glance
> If they ended up maybe changing the wakeword to "the", then maybe they could "silently" listen to everything, but that would be caught pretty quick because the device would be "lit up" constantly (another _hardware_ thing)
From pure technical perspective, can the device not be programmed to be waked by wakewaord “the” with the light off?
I get that you believe this, and I even understand you repeating it to other people on the Internet. What I don't get is that your tone indicates that you are offended people don't believe what you believe... which also just happens to have been incorrect in the past and many others seem to think is provably possible in the future.
> Packet sniffing and hardware inspection both instantly disprove...
I'm under the impression that packet sniffing is useless with end-to-end encryption, but I could be wrong. I.e., you can tell that something is being sent, but you can't know what.
I do understand that the wake word processing happens in a special kernel in a low power state, however....
The wake word is a trained kernel, it can be trained to listen for a huge set of things (as seen in the Pixel's passive song detection), so they would just train the kernel to detect a large targeted (marketing?) vocab.
about being "lit up" constantly? I'm not saying you are wrong, but I'd really like to see a citation that this is true. Is it true for both echo and home?
And while Packet Sniffing can disprove that it's listening and sending whole audio to the cloud, it can't disprove that it's listening for a huge set of "wake words" and toggling bits in other control messages to track users in more subtle ways.
Citation needed. Further, listening for a wake word and reacting to that is likely done completely in software: the fact it's even listening for a "wake word" means the hardware (microphone) is in fact always listening, it's just [presumably] not actually sending that audio to The Cloud (tm).
I don't own either of them but Siri and Google on my phone both require training when I first use them. Do these devices not? IF they do then isn't that proof they are re-programmable and could be programmed to respond to anything?
Not that this helps anyone sleep easier, but imagine in today's age... a whistleblower -- perhaps one of the thousands of software devs working on one of these -- leaked proof that these devices are recording everything to re-market and profit, without permission...
The resulting backlash and legal ramifications would be so huge it just wouldn't be worth it. It wouldn't just take an insane and stupid CEO to do that, but also thousands of other tech/adops employees who'd have to be like, "yea this is a great idea."
Surely somebody in the '90s said something similar with regard to location data, and yet your location is tracked 24/7 by adtech megacorps, and the thousands of tech/adops employees don't say a peep. The playbook has 3 easy steps:
1. Get people addicted to technology X.
2. Keep bugging people using technology X to surrender their privacy using classical dark patterns.
3. Profit!
There is no need for whistleblowers. It's all done in the open. You have already willingly surrendered your communications, your 24/7 location, your knowledge searches, your financial transactions, your media interests and your genetic material. Why not surrender the privacy of your home as well? Yes/AskMeLater.
That’s my big concern with this tech, training people to have always-on surveillance in their homes without a second thought. I realize that the typical and trite response by some involves throwing away my phone, but there are holes in that. First, it is trivially easy to control where your phone is, you can get burners, root your phone, and all of the other good things we know and love.
An Echo, or similar dross is a closed box controlled OTA, and networked. Even if someone had immense faith in company X, it would be unwise to ignore intelligence and law enforcement both foreign and domestic wanting access. You can’t root Alexa, it won’t even work without the cloud. It really does feel like training wheels for something entirely unpleasant, and all because people are so helpless in the face of dubious convenience and fashion.
> training people to have always-on surveillance in their homes without a second thought
Even worse: when always-on surveillance devices become popular enough that a judge could rule that the technology (in the abstract, not a specific product) is "in general public use"[2] - crossing the bright-line rule created in Kyllo v United States[1] - the police no longe4r need a warrant to use the technology see the "details of a private home that would previously have been unknowable without physical intrusion"[3].
I'm not talking about the police being involved with Amazon or using the Echo. When a technology is "in general public use", the police can use their own always-on microphone to transmit previously-private speech to a 3rd party on the internet. Normalizing surveillance devices not harms the person using the device, it also reduces *everyone's 4th Amendment protection.
> Surely somebody in the '90s said something similar with regard to location data, and yet your location is tracked 24/7
I remember a Romanian politician and member of Parliament complaining about the local telecom providers displaying the GSM location data on the phones’ screens sometime back in 2002 and 2003, I remember of laughing at his ludicrous (that’s how I viewed it at the time) complaint, I mean, he was a stupid politician while I was a CS student, couldn’t he see how cool it was to see your neighborhood name on your Nokia 3110’s screen? Of course that the stupid politician was right and I and the fellow technophiles like myself were wrong.
I'm reminded of the Volkswagen diesel emissions scandal, where VW were doing something illegal and were whistle-blown by a developer, costing them billions of USD in fines and massive damage to their brand.
Just because something is ultra high risk, stupid, illegal and abuses consumers isn't apparently enough of a reason for large corporates not to do it.
I just kinda doubt this. How much backlash was there when it came out that the NSA was recording the full content of every cell phone call in the Bahamas?
1) whistleblowing unlikely because any employee that steps out of line can and will be destroyed
2) any media fuss will blow over in a few days
3) promotions and bonuses require outsize risks
I think everyone has a point in their career when they realise large tech companies are unaccountable before the law. Mine was watching the MERS database running roughshod over American property ownership laws.
> The resulting backlash and legal ramifications would be so huge it just wouldn't be worth it.
Everything can be explained away with "we discovered a bug that might cause your unit to record you constantly, but it's fixed now. Won't happen again, sorry!"
> However, Apple's practice of sharing Siri data with third parties [to provide and improve Siri, Dictation, and dictation functionality] is perfectly legal and outlined in Apple's iOS Software License Agreement, which Siri users are required to accept.
I mentioned using the voice data to market "without permission." That's my rationale. All of these scary location tracking this, retargeting that methods are always buried in a privacy policy somewhere. But when you start doing it on the DL, that's when you get in trouble. So the cons greatly outweigh the pros for any sane company.
And of course the second article is an isolated case of human error. Nothing to do with violating privacy for profit.
"Allow Google Maps to always access your location. Yes. Ask Me Later".
Given the multiple precedents on the erosion of privacy path in the past 20 years, of which I quoted one example above, it's pretty obvious that they will turn "always on listening" in the future, using whatever dark patterns necessary to avoid a class action suit.
Because that would be pointless. The real problem right now is false triggering. If you’re actually worried about being spied on, why on earth would you have one of these in the first place?
From a non-privacy perspective this adds the feature of being able to customize your wake word, which besides being a nice feature on its own also counters Google and Amazon's desire to inject their brands deeper into our psyche by making us say them out loud.
From a privacy perspective, having user control of the wake word prevents Google and Amazon from adding future wake words that could be abused for other ways to track us. For example, Google might get the bright idea to track TV ads by listening for audio in the ads. Or tracking people in your house by making Android phones emmit non-audible chirps. These kind of "features" could be easily introduced at any point in the future by an update to privacy policies that nobody notices.
Google might get the bright idea to track TV ads by listening for audio in the ads
Considering that some TVs have this built-in, I'd be surprised if Google wasn't already doing the same. It's why my "smart" TV isn't allowed to connect to my wifi network.
(There was a previous HN article about it, I believe the brand was Samsung.)
I think it was the Facebook app, it listened to ambient background to determine music and TV shows. The user had to opt in or at least approve permissions.
Afaik by now, there have been least two court cases where Echo recordings were handed over as evidence [0] [1].
At this rate, it's only a matter of time before evidence like that gets leaked/released, which would serve as a good probe on how much these devices really record.
Who cares. Whoever buys the hardware can do whatever they want with it. If someone feels better with this device on their Google / Alexa product let them do it. If the speech recognition is horrible I’m sure they’ll take it off.
> If the speech recognition is horrible I’m sure they’ll take it off.
This is interesting and cool, cause it sounds to me like they only detect for the keyword to "unlock" the Google home, so in theory they don't even need speech recognition. In most cases, they could just do with telling if a sound you make seems to match the sound you defined to be their name ¯\_(ツ)_/¯
Also, it would make no business sense to always be listening. A lot of people think that just because Alexa and Google Assistant are free to use, it means that these services are virtually free for the companies as well, but that's not the case. There is no way Amazon or Google would waste millions of dollars running the state of the art speech recognition algorithms on your house's background noise.
There is no way Amazon or Google would waste millions of dollars running the state of the art speech recognition algorithms on your house's background noise.
Unless they wanted to listen for a dog barking, then add "pet food buyer" to your profile data.
There are a lot of very easy use cases for monitoring background noise.
> Users without the skills to verify the code isn't nefarious have to trust good samaritan developers instead.
I trust that amongst thousands of people with different incentives at least one will raise their voice if something is not right. At least more so than I trust a corporation with, in this case, the the wrong incentives to self-regulate to my expectations.
Nothing is 100% guaranteed, but with an open source project, given enough users, its far less likely for someone to be able to bury nefarious stuff without many eyes looking at it and at least one person sounding an alert.
Yeah but really this isn't true. Popular open source that has tens of thousands of eyes on it still gets compromised all the time (see: npm). Even the Linux kernel has had rogue git commits injected into it.
The probem with npm isn't that open source doesn't help, its that the eyes get spread out thin when you have thousands of modules - so nobody is looking at the changes that happen in their lots of small dependencies.
Which is not to say that thats not a valid approach - but for it to work we need better tools to handle lots of git repos at once (for example, the ability to get notified about any new code on github that affects your project would be pretty cool, especially if its coming from people or organisations you haven't explicitly marked as trusted yet)
I would like to see someone try and sneak rogue commits into Linux. It would be quite the feat.
This is code you can inspect running on hardware that you own and control. It's trivial to ensure it's secure at that point. Unlike when it belongs to a company.
It doesn't have to connect to the internet to do what it does. The scenario you seem to be suggesting is that the Project Alias developers would be conspiring with Google by compromising Project Alias to NOT disrupt Google's listening and then Google would be listening in on you using their network access. This by definition does not double the area of risk.
If you can be confident that Project Alias does not have network access, then the worst possible scenario, even if the developers are literally Satan, is that Google Home would be doing exactly what it does without Project Alias attached.
He's not talking about a rasberry pi in general, he's talking about Project Alias, the device featured in this article, and the first step in the instructions is connecting the Pi to your Wifi so you can download the software.
So yeah, this project turns the raspberry pi into an internet connected listening device.
Also says in the same instructable that once the device is trained that there's no need to have the device connected anymore. It also doesn't need to be connected to the internet - you just need to be able to get to it via a browser with a microphone - I was able to train the device with no connection to the internet.
It's probably the same reason that people started installing listening devices in their homes in the first place -- convenience. Few people want to walk around and plug in a cable to update/reconfigure their devices.
Edit: It’s very unlikely because Amazon and Google pay for false positives, so they have a strong incentive to develop really good trigger word detection.
If the false positives give them data that increases the value of your marketing profile by more than the cost of processing the interactions then they actively profit from false positives, and have no reason at all to stop them.
The point is establishing as much trust as you can with your devices. Only having external points of trust when absolutely necessary. This is generally a Good Thing, and something we should do by default, not as a reaction to some corporate leak.
The sentiment this is countering is not "Google and Amazon _are_ recording 24/7/365". It is countering "Amazon and Google _would_ record 24/7/365 if they could get away with it socially".
Arguing that the technology doesn't do this does not address the underlying root perception that Amazon and Google are not to be trusted.
Google built its entire business on harvesting _all_ data on the web and building an infrastructure to process it efficiently. Purchasing Nest made clear that to us that they now wish to harvest data from the home. If they were able, socially and politically, to harvest _all_ data from the home, their history indicates that they would do so with all possible speed.
Having a device you built yourself that learns locally and is under your control prevents Amazon and Google from changing their mind about what level of recording is acceptable without your informed opt-in consent at the time of the change. Both providers reserve the right to update their privacy policies without notification or consent, including granting themselves to increase data collection.
TLDR: This device is the physical expression of mistrust of Amazon and Google. Their greed for your metadata is well-documented, and their policies let them increase data collection in your home at any time without your consent. Such increases would be prevented by this device.
> While I appreciate the sentiment...unless you actually think Google and Amazon devices are recording irrelevant ambient sound deliberately (they aren’t)
Citation needed.
Or if you have reference firmware I can load, that would be awesome... What do you mean its closed firmware and controlled by Amazon/Google? You mean they can change it whenever they wish, and we have no say???
Long story short; you rented a spy device and you trust some random person online it isn't spying... Even though there are credible stories of these devices doing precisely that.
Every cellphone in the world has a microphone that could be listening all the time and sending data anywhere. So does most every computer. It's a better threat vector by 1000x, more stealthy, easier to conceal traffic. But all anyone ever talks about is a device designed to listen to you talk because hey, so obvious, big brother MUST be listening in there!
Cell phones have batteries, so it would be even less practical for phones to be "phoning home" a stream of what's going on around it at all times than a "smart speaker".
Here’s a parallel construct: It would be trivial for Microsoft to have a key logger in Windows that sent every keystroke to Redmond.
Why does anyone trust that they are not doing it?
Indeed, they could scan your computer for whatever data they like and send it to Redmond. But no one even suspects it, or at least at nowhere near the level people seem to distrust Amazon and Google over sending your voice to them full-time.
This is what I mean by Misplaced Distrust. We already trust Google and Amazon with far more of our personal data without a thought.
And metered data. I would be very aware if my phone was eating my data plan via constantly recording audio. Even a measly 12kbps audio stream adds up to nearly 4GB/mo if recording 24hrs/day.
Non-techie users would absolutely notice that their data and battery is being used up in the background pretty quickly. Further, the cell networks simply could not support that kind of usage from every single subscriber at the same time.
Your phone could easily wait until it's on wifi to upload 24hrs worthy of ambient voice data (which is only trigged by voice activation, so not a full 24hrs)...
The feds have been using cellphones as full-bore wiretaps since the early 2000s when they used it on mobster's
"dumbphones". I'm sure they've figured out clever exfiltration techniques on smartphones by now.
Especially considering how willing the ISPs/telecom companies are to bend over backwards to hide surveillance. Even 3g/4g wiretapping is probably feasible.
You have a point, but what makes you believe that such things are not happening for some people? Do you have any proof that the low power microphone is not recording voice all the time?
So doesn't Alexa already not record until you say the trigger word? If we don't trust that that is the case, then sure this device covers that, but it doesn't change the fact that they are still collecting data on every command you issue to the device.
That is correct! Google, Amazon, and Apple despite having vast resources don't have the ability/desire to pay for the bandwidth, storage, and processing needed to have 24/7 recording and analysis from every smart device, especially when you realize that includes cell phones.
Also it is ironic because all modern cellphones have the whole smart speaker thing built into them but people aren't freaking out about that in the same way. (Probably because they haven't realized it. ¯\_(ツ)_/¯)
This is similar to the issue of people being scared of radiation from cell towers but somehow not freaking out that they have a transmitter in their pocket that uses the same range of frequencies but at vastly higher power (due to the proximity).
Cell phones don't leak precise power signals that identify exactly what a home user is doing. Smart Electric meters are absolutely part of the surveillance capitalism equation. Power Analysis is a leaky side channel.
Also it is ironic because all modern cellphones have the whole smart speaker thing built into them but people aren't freaking out about that in the same way
Actually, they are.
The majority of "ordinary" people I interact with believe Facebook is listening to their conversations 24/7. It's been brought up multiple times on HN.
>The majority of "ordinary" people I interact with believe Facebook is listening to their conversations 24/7. It's been brought up multiple times on HN.
I see this a bunch when people say, they talked about an item with someone then google showed them an ad. Sure, perhaps all that data is being processed, but more likely, it's predictive and they know who in your social circle knows about this item. Your friend read a story about it, searched for it on amazon.
People are still in this mode where they don't realize how much info leaks without capturing any conversation, and how good prediction is/can be at this point. So the only explanation is that we're being spied on all the time because I don't think people can wrap their minds around prediction models that are this good.
There are stories from 10 years ago of advertising agencies knowing that someone is pregnant before they did and ten years later, the general public assumes they have to have the original thought first so FAAMG knows what to sell them.
>The majority of "ordinary" people I interact with believe Facebook is listening to their conversations 24/7.
Which is to their point because this is more an example of dunning-krugers with technology than anything else. Those "ordinary" people are data dumb, and project their own explanations onto data joining, which in this case they can only explain as 24/7 microphone recording.
That would either increase the cost of the device or degrade the performance of the speech-to-text.
In the case of your phone, you'd get about an hour of battery life. Reasonably accurate speech recognition requires several orders of magnitude more computation than wake word detection.
My experience with pocket dialing and speakerphone calls leads me to believe my phone is a much less capable listening device. Many people still have phones where constant listening would supposedly drain the battery.
Google Photos allows users unlimited storage when storing photos in their proprietary format. I see a 90-day buffer per user consuming roughly the same amount of storage.
Also disingenous to consider the smart speaker that works without cell reception in the same category. Those that require cell reception to work are just as circumspect as Alexa/Google.
The microphone on the Alexa needs to be constantly active so that it can hear its name (the trigger name). The point of this device is to defeat the microphone, by playing white noise.
But yes this device does not address what data is captured after you have engaged your smart assistant (nor is trying to address that issue)
The processing for the wake word is handling on the firmware. If you point Wireshark (or another network monitor) at a Google Home or Alexa device you can see that there isn't significant network activity while idle.
If they (the smart assistant makers) would promote the fact that their devices only go online after the trigger word is detected that would go far is assuaging people's fear of the always-on microphone
The reality is that conspiracy theorists aren't interested in learning how things actually work, which is why they're conspiracy theorists.
In reality your phone is a significantly bigger threat to privacy both in terms of normal day to day monitoring (e.g. location tracking) and even listening in (a closed source baseband that can communicate on a platform you cannot even monitor).
what's crazy, is that one of the biggest leaks here is cell tower location tracking from triangulation. Like not the gps data that gets constantly sent to google/apple but the fact that there is a middle man that sells all that shit to whom ever. This has been a problem since before smart phones but it's mostly transparent to the user so people don't seem to be bothered by it.
When people see the GPS is on all the time or can look up their location data off their phone, it feels uncomfortable, but then they make un-encrypted phone calls through ma bell and let their location get sold.
People are really bad a privacy and key on these obvious, invasive feeling, technologies but then miss the bigger picture.
True, but hopefully that's banal things like "play Despacito" or "What's 2 + 2", not "Please add 'rending me in the gobberwarts with a blurglecruncheon' to my depraved kinks list" or "Set my root password to 'secret123'" or "How do I build a nuke in my basement off of stuff I can order from Amazon".
We do indeed throw horse porn into eachothers' search history. I sense your sarcasm - perhaps my friends and I have a different group culture than yours do?
Maybe, but the system is designed in such a way that a third-party could listen to the convos anyways, even if Amazon/Google choose not to do it (for fear of backlash, and no other reason):
This is a brilliant idea. I'm curious about a lot of the implementation specifics like how audible the white noise is in a silent room and what kind of UX trade offs it brings.
Yeah, I'm wondering how it should work when you say the wake word too. Like if I say "Sister Assumpta, what's the weather like?" is it just going to repeat the same command but replace "Sister Assumpta" with "Hey Google?" Thus slowing the whole thing down?
Seems to me based on the description that it only says "Hey Google" to the device and then lets your voice through by cutting off the white noise. That would mean the only delay would be between when you say the wake word and when Google starts listening.
The article says that after your keyword it deactivates the white noise wall for a while, and whispers the original keyword. So alexa / google / siri will hear you directly.
I'm curious whether it will handle the audio trigger as well as Google/Amazon/Apple. It takes some pretty advanced audio algorithms to catch "hey Google" in a busy kitchen. It would be pretty frustrating if this device caused these devices to trigger even more inconsistently.
Ya, my echo devices can hear me across the kitchen, with people talking near it, while the water is running, and the TV is on (in the distance). That's a core feature for me. If this device dropped that capability, I'd rip it off in a day.
Nope. The Amazon Echo uses an array of seven microphones to isolate speech from background noise. There's no way that this device can pass through accurate spatial information, which will inevitably degrade speech and wake word recognition in noisy or reverberant environments.
Everything about this is radical aside from the way they made it resemble an actual fungus growing on a device. This is the most extreme skeuomorphism I’ve ever seen.
You could also just not buy one of those awful things. I have never seen a legitimate use for it that wasn't misplaced adolescent tech fantasies (omg I can tell big brother to make coffee and my keurig starts up!). But maybe my line of business has made me excessively paranoid / niche
I would like to make an edit: functionality for those with disabilities is a huge use-case I did not consider. Thank you for your insightful comments
I like to be able to play music, ask simple questions, etc. without pulling out my phone. I don't understand why those aren't "legitimate" use cases. Maybe you don't like the tradeoff you're making in using such a device, but if I'm fine with it, how are my uses cases not legitimate?
Yeah, I got one for Christmas a couple of years ago; thought it would be really gimmicky, but I find it's actually quite a lot nicer to use than my phone or other devices for the following:
* Turning on/off lights
* Changing the thermostat set temp
* Asking about the weather forecast
* Add items to a shopping list
* Playing music
If you can't imagine a voice interface being appreciably better than a phone interface for these things, I recommend withholding judgment until you've tried it.
Also, to be very clear (since some people have a knack for arguing points I never made), I'm explicitly not saying that the current hot-mic implementation is technically necessary or ideal, nor that these conveniences justify the privacy tradeoff.
It's also worth noting that you don't always have to do it with voice either. If I'm on the couch with my phone, I might use it to update my shopping list, whereas if I'm in the kitchen cooking, I'll use voice. Similarly, if I just want random music, I will say "play music", but if I want a specific album, I might use my phone, especially if it has a complicated name that I don't remember.
Obviously not every action is easier or more optimal by voice, but having the option is great.
Apologies if this comes across as too "get off my lawn", but I come from a time when to look something up, you had to haul yourself to the library; open one of dozens of drawers filled with index cards; find the card your looking for, which directed you to a stack in the library; find the book on the stack; and finally find the page in the book by consulting an index. It's a lost art. Then, you would have to go to an actual person and engage with them in order to take the book home with you, giving you time-limited access to the information. That's if you're lucky enough that the book existed at your library. If it was checked out or had to be ordered, it might take weeks for you to get access to that information.
Many people today grew up with cellphones in their cribs. They have no idea what information starvation is like. The experience of receiving information you've been waiting for for weeks or months is exhilarating.
Anyway, when you take the library experience versus the experience of pulling up information from a cellphone the improvement is astronomical. From cellphone to voice assistant, the improvement seems very marginal.
Cell phones even represented a distinct advantage over desktops and laptops in that they were always there on your person. Cell phones opened up the possibility to look up information anywhere. With voice assistants it seems the only advantage you gain over cell phones in that you don't have to use your fingers. That doesn't seem very life changing by comparison, unless you don't have fingers, in which case I will admit your life would be vastly improved.
But the downside is that you're connecting a always-on microphone access to mega-corporations who are looking to monetize your existence. For those of us who grew up without the internet or cell phones the trade-off just makes zero sense. We're willing to use cell phones because they open up new worlds of information access. But voice assistants just seems to create more problems than they solve.
I too grew up in the time of card catalogs. And I learned a lot from reading through the other encyclopedia entries as I flipped the pages looking for the page with the info.
Yes, you're right, the voice interface is not the astronomical leap that the cellphone was. But why is that your cutoff line?
My voice assistants offer a lot of benefit to me. Especially with kids, I don't always have a free hand to pull out the cell phone. When my daughter was an infant, it was super convenient to ask it to play soft music as she was falling asleep without having to put her down. Now it's super nice to be able to set multiple timers as I cook with just my voice, instead of trying to fumble with multiple timers on my phone or stove.
I'm not paranoid to think that they are recording everything, because I understand that there would be no ROI for the company to do so with the storage and bandwidth that would be required. And therefore there really isn't much tradeoff at all. Google is already recording every search I do -- does it matter if I use my phone or my Google Home?
My personal experience is that simply typing my query into a search engine or pressing the spotify logo to start my music requires less effort or fuss than attempting to figure out how I'm supposed to word my desire for the benevolent overseer to do what I want.
IE, using voice commands is a downgrade IMO. Voice commands are not directly discoverable, and there's a lot more magic boxes.
> attempting to figure out how I'm supposed to word my desire for the benevolent overseer to do what I want
I have yet to find a use case for modern voice control that required more than a passing thought about how to word things. Even my technologically illiterate parents can use these devices with relative ease, especially compared to smart phone and desktop computer UIs. Have you actually tried out these devices or are you just assuming they're as bad as they were 20 years ago?
I thought there'd be more problems with voice commands, but the Echo's not bad at it. I have to repeat maybe one thing a week, but "Echo, [room] lights on/off/dim to 50%" always works, as does "play X". I don't find myself having to structure a command in any particular way.
I do find using voice commands a downgrade when it comes to interaction speed. I find it incredibly annoying to talk to alexa as it doesn't seem to match my dialog speed. Then, I find myself standing their waiting for it to shut up thinking, 'I could have done this faster myself'
Also, an interaction I had last week:
add x to my shopping list.
ok, I will add x to my shopping list, anything else?
. . .
But I can't add a list, add pears, apples, and oatmeal to my shopping list.
So If I have raw chicken on my hands and want to add shit to my list, it takes so god damn long that I want to punch the fucking thing.
I don't have an echo but there was a post on HN[0] not long ago that linked to an article[1] about how Alexa is able to add multiple items to your shopping list at once, and how it understands what is what. So in theory you shouldn't be running in to the problem you're describing. Not every time at least.
I'm sort of on the boundary there. I didn't have a cell phone until I moved away from home for school. I had occasional access to dial up Internet as a kid, but I still remember plenty of afternoons spent at the library.
Rarely do I need to answer a simplistic question in line with "what's the time now".
Its rather some deeper info that first few lines of wikipedia article covers, sometimes more. Is your use case valid? For sure for you. But it wouldn't be enough for me, not for the price, upfront and hidden, not for the creepiness it potentially brings. The real time and energy saved for me would be tiny - but that's me. I can still do a bit of 'work' myself.
This reminds me of my recent trip to Aconcagua, highest peak in South (both) America. One US lady had this electric air mattress inflater, and she ran off the charge. She was bragging how smart is she for having such appliances. Rest of the group just smiled and inflated our mattress ourselves, even in 6000m high camp. If you can't do 10 full lungs blows yourself and spend that 1 minute preparing mattress, you shouldn't be up there, by huge margin.
I have four. I like being able to ask a question while I'm still typing away. I like being able to turn on my TV and change the volume without having to find a remote control that is always somehow somewhere it shouldn't be. I like being able to turn off every light in the house and the iron by just saying "alexa, turn off everything" as I head out the door. My son loves being able to turn off the lights in his room and start a playlist to help him go to sleep with "alexa, good night". I love being able to start any playlist from anywhere in the house without touching any buttons, or turning the lights on or off when my hands are full without putting things down. Or asking for the time. Or when the next bus leaves the nearest bus stop.
Any one of those things are just a tiny little convenience, but it adds up, and while I bought one just to see what they'd be like not expecting to use it that much, I now use them dozens of times a day.
I'm probably biased because I lived my early childhood behind the Iron Curtain but I can't for the life of me understand why someone would buy these. The cost-benefit is just not there.
the cost is there :D just no benefits, like most 'products' today... i can't imagine 1 unit being sold to anyone but some addicted person who needs to have all-the-new-things regardless of their use.
I agree that your phone can delegate the same voice commands to your smart devices or do Q&A, but some of these devices are also decent speakers to stream music casually (eg. $75 Echo).
My phone is in my pocket, which signficantly degrades the audio quality of any recordings. Same reason I have a cover over my laptop camera, but not over my phone camera.
I've made several audio recordings with my phone in my pocket. Unless you're very consistent with the pocket and placement of your phone, that isn't a significant mitigation.
Huh. I came to the opposite conclusion - growing up in Eastern Europe, "they're always listening" is the default position for me, so one more listening device is no big deal.
Do you have kids? I have three small ones, and they are just starting to desire technology. From my perspective, letting them control music (which they want, and I want them to have) is much better using a Google Home device than giving them access to my phone or tablet.
If you don't have kids, you have no idea how loud and aggressively they will scream when they want something, and especially when these devices are visible to them (if you "need" to respond to a text message, etc).
Yes, it is a devil's bargain. Yes, I'm sure some families are able to, through sheer force of will, completely restrict access to technology. In my family, we are acknowledging we have lost the battle to prevent them from using technology and are seeking solutions that help them manage their desires and create healthy boundaries. I guess we can all argue over what is "healthy" and "normal."
Things like Google Home and Family Link (all from Google) do allow us to control access in a way that I prefer.
So, this "hack" is really exciting because I do care that my two year old already knows Google as a brand.
I'm open to hearing suggestions and have even attempted to build my own open source alternatives, but using voice is a modality that is preferable for so many reasons, and I don't see alternatives that won't be worse.
I have two kids, 9 and 11, and neither of them are screaming and yelling about what music to play. If they want to listen to music, they know to go to the rumpus room and pick out a vinyl.
It's not a devil's bargain, you have lost the ability to bargain. It's a common theme with parents these days.
There is nothing magical about technology, it's just an application of age-old parenting principles. And there's nothing particularly harmful about technology either, there should be no grand battle: you define the limits and the children should stick to them and respect you as a parent. This is true for all things children want to do, from screaming and playing indoors to accessing communal devices to getting their own devices when you, as a parent determine they should.
You are wrong. With kids, you are dealing with micro bargaining every moment. It's what kids do to learn.
Do you have kids, or did you read this somewhere?
What age old parenting techniques are you talking about? The ones older people reminisce about when they lament how bad young people are today? Do you have a source backing up the efficacy of those "proven" techniques?
An aggressive screaming kid needs a timeout at the very least, followed by a progressive loss of privileges (toys) until the tantrum subsides. A few cycles is enough to amend even the most recalcitrant.
I'm shocked: Why does your two year old need to know Google as a brand? How or why is this valuable to you? Do you expect Google to exist forever? Its entire revenue model is built on ads. Companies with more robust revenue streams have gone bankrupt in shorter timeframes.
Are you sure that gets what you want? Our desires for my kids might be different. Sounds like you think kids should be punished until they learn who is the boss. I'm not sure you have read all the literature on the effectiveness of that strategy.
I never said I want my two year old to know the Google brand. She hears her older siblings saying it. It is just what is so with her. But guess what? I'm willing to wager my kids aren't the only ones who learned things from their siblings that their parents don't want them to know about, at least at that moment. My kids are not playing with Barbies and I'm pretty sure body image issues with girls are much worse than exposure to Daniel Tiger.
Sorry, my apologies. I misread your post read to mean that you were happy/excited to have your kid understand Google as a brand (i.e. valuable).
As for parenting, we may just agree to disagree. I concur with your assessment that siblings will definitely teach more than parents. That's to be expected. We just would never reward bad behaviour with acquiescence. But to each his own. Our seven-year old has wide latitude when it comes to choices and actions, but he also realizes that the consequences of those actions are not in his control. We gave him his own iPad at the age of 3 and access to his own real spending money in Grade 1. He gets to decide what to spend it on. At the same time, we've made it clear to him that poor impulse control and bad behaviour will never get him what he wants. He negotiates everything, including daily bedtime or routine tasks, and we're perfectly fine with that. It seems to align well with his personality, and builds some valuable life skills.
I actually think we agree on more than we disagree.
Totally align with not giving in to screaming and yelling, and I'm consistent (or at least aware) about that, but when my youngest is sick and just went down for a nap, well...
Those are good points you make and I'll hope to recall those techniques with my just turned six year old. As you say, building valuable life skills.
Some may see smart devices as part of a greater cultural movement, in which corporations entice individuals to trade privacy for convenience
Someone who thinks these things are a gimmick, and a harmful gimmick at that, is right to express an opinion about the value these devices add. The same way I'd encourage a friend to quit chain smoking tobacco cigarettes, and not visit his house with my family if it were full of secondhand smoke.
Peer pressure is the main reason I quit smoking. So yes-- if not every time, then at least regularly and consistently. When technology starts to look like a recreational drug, treat it like a recreational drug.
Why is this downvoted? Peer pressure is by and large the best way to curb thoughts and actions that are harmful to individuals and society. The current echo chamber on the net filled with alternative/radical theories is the outcome of insufficient peer pressure. Nutbars always existed in real-life too, society was just better at keeping them from doing too much harm.
Those of us who are nutbars hate society for forcing us to conform. Given that a lot of us are more tech-savvy than society, we love that we tend to be able to work around its restrictions.
I think it's dangerous to conflate freedom of speech with moral relativism. Freedom to speak does not suddenly make morally wrong actions right. Ex: murder, slavery, forcible confinement, poor treatment of women and children, and so on.
I realize we are quibbling the definition of "nutbar", but my line is drawn at the extreme end, not the moderate end. It's one thing to advocate Haskell as the perfect language for building an OS (crazy talk, but I support it) and quite another to advocate violent uprising against minorities in society (a la StormFront). I hope we can agree that there's a distinction at play here.
It's somewhat poignant that you say "adolescent" because I know a number of octogenarians and nonagenarians for whom these are indispensable. I won't buy one at this phase in my life, but there will come a point in most people's lives where having an omnipresent corporation listening in will become a net positive.
I would of course prefer dear friends listening in, but who knows what life will bring.
Never seen a single use case? How about a kitchen assistant, especially with dirty hands, or a task doer for the elderly or otherwise mobility disabled. It’s a generic time saver, and time=money for most people.
I mostly agree with you but there is a legitimate use.
Imagine if you couldn't use your hands or interact with technology due to a disability - these devices would make the world so much easier to interact with.
Anyone who grew up watching Star Trek has probably dreamt at some point of being able to simply say "Computer" and then send a voice command. We just didn't expect the tech to become popularized by advertising/retail companies.
My dad just had some major health issues, and has a much harder time getting around the house. He finds these things pretty useful to save him some trips around the house.
For one, Amazon doesn't make smartphones, and I trust them the least.
In the iOS ecosystem, Siri can be set up to only listen in response to a button-press (don't remember which is default), and the watch only listens on wrist-raise. Those at least creates some physical barrier to passive listening, even if it requires a certain degree of trust in the devices.
> In the iOS ecosystem, Siri can be set up to only listen in response to a button-press
Sure, that's assuming you trust Apple (if you don't, then you might imagine that those settings don't actually do what they say they do). If you trust them, then you're fine. If you're in Google's ecosystem, and you trust Google, then you're fine. Ditto for Amazon. But the point here is that people don't trust these entities.
And then somehow target a smart home speaker, while simultaneously carrying around an always-on, always-connected, geo-located, potential listening device in their pocket, all day long.
I'm privacy conscious and bought a couple of minis. I don't use the mics but I thought I had a legitimate use case without one. It's falling short and I'm looking around for replacement devices.
I flipped the hardware mic switch off on it, to try to make it a dumb wifi speaker instead of a "smart" one. Then I built a software alarm clock that forces me to leave the room after I wake up in order to turn it off.
For me, it's very important for my alarm clock to be both effective, and to always work. The alarm clock runs as a remote task and connects directly to the mini, telling it to play an MP3 from the local network (by IP because the mini ignores DHCP DNS server). If I hit the mini touch controls to turn it off, the software starts playing a different MP3 a second later. If I try to unplug the cable from the device, duct tape stops me. Thoughtful wrapping of the cable around a solid furniture post prevents any yanking from being effective at tearing it out of the wall. If one mini is down (fairly rare but possible point of failure), the other one is attempted.
So, it's fairly impossible for me to just turn it off without waking up and giving it a bit more thought. I have to leave the room and tap a button on a touch screen (ubuntu in kiosk mode reaching web app on local network).
The unfortunately fatal flaw is that after months of effective use, I recently discovered that my highly available alarm clock was not actually highly available. It breaks when the internet is out. I could not connect over local network. There's always the possibility that something else was a factor, but I reproduced it a couple of times intentionally.
It also concerns me that the mini doesn't require authentication. Anyone on the local network can directly reach the device and do the same thing. A script meant as an alarm clock could turn into a device of psychological torment in someone elses hands. This lack of authentication, and the ability to auto-discover the speakers, is probably something they consider a 'feature'. I don't like seeing Chrome waste system resources in its attempt to scan my local network on the off chance that Google's speakers are there. And I don't want it to reach out to those speakers when it does find them. But it does it anyway.
In the end, with the microphone disregarded, it's a cheap wifi speaker. I won't count Chrome's bad behavior against it, but its software could be improved by offering (any) secure connection options. The lack of internet as a single point of failure dooms any kind of gadgetry with a reliability requirement from using it. It can't be considered reliable enough for serious tasks like waking you up for work or a flight unless they fix the software to work in a local-network-only mode. But, it is cheap, and, well, mostly available, which is often good enough for to-hand use cases.
Speculation: Is the lack of mini's heartbeat phoning home Google's own way of determining network reliability across wide geographic areas (eg, the lack of data in an aggregate area)? But they probably know this already from the wide spread of Android devices. Or do they maybe just not want their device to work unless it can reach back to them?
I love this idea. I've wanted to do something similar for a long time, but I was thinking about building a home assistant where the always-on mic was a complete separate board that only listened for wake words and had no internet access. The main mic would only be powered on when the smaller board woke it up. Alias achieves the same thing in a much simpler way to where I might actually consider buying a home assistant now.
How is this different than an actual echo device as it sits now?
You're still relying on a piece of software to make the wake-word assessment and hand off the audio to the cloud. Now you're just adding more hardware parts to fail.
If your argument is that you trust your software more than Amazon's, then you shouldn't need anything more than a single microphone anyways because why would you surveil yourself?
> ... you shouldn't need anything more than a single microphone anyways because why would you surveil yourself?
For the same reason nginx usually runs as a separate user: people make mistakes and security vulnerabilities happen. Security in depth is a good thing.
Mostly what supermatt said -- this is a lot simpler. Particularly, it's been a long time since I looked, but when I last looked neither Google nor Amazon were offering the voice assistants in nice package where you could bring your own hardware but use their APIs. So beyond the hardware, the project would've also involved building my own voice assistant software as well. Both of those ecosystems have evolved alot so that might not be a hurdle anymore.
Like hey, let's install some under developed AI from some unknown company with unknown security policies on top of a device with access to hordes of personal data and the ability to make transactions online.
No hacker will EVER think to use it as an attack vector /s
I'm much less interested in gating audio recording (which I have reasonable confidence in the device itself doing) and much more interested in being able to use a device like this without turning on all of the various histories. Google Assistant refuses to do most of its interesting functions (other than trivial things like setting an alarm) without turning on search history, location history, voice history, and various device information.
There's no good reason for it to require that information for a request like like "play XYZ on YouTube".
> which I have reasonable confidence in the device itself doing
A few horror stories related to Alex hint that it might not be doing a very good job. The grammar/syntax it uses to wake is much more complex than what Alias is proposing as a safe alternative. The most blatant example would be the Portland, OR couple that found the Alexa device making phone calls to people as they had a discussion near it.
I don't know about Alexa but Google Home always repeats back to me in a loud and clear (and slow :( ) voice what it's about to do when I issue a command. I have a hard time believing you wouldn't notice the device making a call. Even if you failed to notice what the device was doing, the likelihood of this type of mistake vs. the likelihood of a pocket dial seems relatively slim.
Do Alex and Google have software APIs? Can you make a hardware device (eg Raspberry Pi) that listens for a wake up then sends to the API? Seems more elegant.
As a bonus, maybe your device could understand "Alexa..." and "OK Google..." and send to the relevant API. Use Alexa for shopping and Google for searches?
Google assistant has an API here [0], and Alexa has a fairly large amount of documentation here [1]. In theory, the devices already claim to do this. If you don't trust that they do, you probably shouldn't have one full stop.
Hm, I don't quite understand how this works. So, if you say "Alias", then Alias has a built-in speaker that whispers "Hey Google" (or the equivalent) into the microphone. What if I start the question before Alias has finished analyzing my speech and playing the "Hey Google"?
Assuming they didn't do that already, a couple more mics in the device base with noise cancelling circuitry if implemented the analog way or in software will defeat this project in no time.
I applaud the intent, but unless one knows for sure how many mics are there, including tiny MEMS devices like accelerometers that can be used as such, and can physically disconnect them all, there's no way to know the device isn't listening.
Scary stuff, don't read HN past bedtime! It's probably more tricky to pull of this trick with complicated brains, but who really knows how we tick and who is running us.
Access to privacy equals business today.
Therefore turning this "promise of privacy" project over to the business sector would be a contradiction in terms.
Most of the world does not care that companies/governments have access to their privacy.
Snowden is a good example of this principle.
The few that do care would not be enough to sustain this "promise of privacy" business model.
Therefore the responsibility is upon each individual to make their own Alias in order to protect their own privacy using this open source maker project. Great work to all those that worked on the Alias project!
This is going to mess up the voice detection. It's keyed off of the original trigger word and knows who is talking to it so you get reminders and other items specific to that person.
Well for one you would need to build user identification into the speech recognition on Alias which is likely a lot harder then just recognizing one trained trigger word. Could be a feature down the line though.
That was my first thought too, you would have to train the device to use your voice sample for the whisper of "hey google", which is totally doable. My main concern is speed.
That is, use Google Home/Alexa/Apple Home for their speech recognition abilities while ensuring it doesn't eavesdrop and works with any other smart device that you have (e.g. there is no way to directly control Nest using Apple Home).
There's really no need for that. Google at least has a speech recognition API that you could send audio to directly; no need to use a Google Home as a middleman if that's all you want.
“Echo woke up due to a word in background conversation sounding like “Alexa.” Then, the subsequent conversation was heard as a “send message” request. At which point, Alexa said out loud “To whom?” At which point, the background conversation was interpreted as a name in the customers contact list. Alexa then asked out loud, “[contact name], right?” Alexa then interpreted background conversation as “right”. As unlikely as this string of events is, we are evaluating options to make this case even less likely.”
"For these instances, Amazon claims that the devices were likely triggered by false positive commands."
in my mind, it's not a question of if the device is recording you, because that's exactly what it's made for. it's a question of, if the company or even worse, govt. want to use these devices for spying or info gathering.
we already know the NSA has back doors and exploits they explicitly decide to keep open so they can access devices for information gathering i.e spying.
Perhaps someone will build a little robot that listens like this Alias, then takes the Alexa/Home out of a soundproof box (or just opens the box) and passes on the commands, then closes the box again. That'd get rid of the static...
From my understanding, just because these devices are always listening, doesn't mean they are always recording. Only when you say the wake word do they start recording and only that information is used for advertising/etc.
It’s a closed source device with an always on internet connection and it silently accepts software updates. So... maybe? Certainly that is how they are marketed. But is that really how they behave? Does the software accept a command that would have it send back additional data on command? Could we know?
I bet Echo and Home could be reprogrammed by Amazon/Google to listen all the time, learn what the Alias trigger word is, and speak/replay that word for the Alias shell whenever they feel like.
That's not how this works. Alias plays white noise into the Home/Echo speaker so it can't hear what's going on, unless you first speak the Alias' wake word, which causes it to activate the Home/Echo and allow speech through.
This is an extremely clever solution to the trust question for people that trust the company with their query data but not with every snippet of raw audio that could float through their room.
So the market for this device is.. people who care enough about privacy to buy this device, but not enough to forgo using an Echo or Google Home in the first place?
In the “let’s go down the what could be done” path...
No network accesss doesn’t mean no access to the outside world. “Alexa, tell some tracker that ...” when you are asleep. And how many people audit their past activations on the Alexa app?
No, I don’t believe this would be the case... I have half a dozen Alexa devices of various builds, a google home (that is currently powered off) and some iDevices plugged in for hey Siri in different rooms.
The thing I was trying to point out though is that no network access is not no outside world access for this device. And if one is paranoid enough to desire this device, then the device itself should be worried about.
And yes, with the right software, it could be reprogrammed via voice too. “Alexa, read some reprogramming site”.
The only acceptable solution is to make it yourself and audit the code yourself if you are concerned enough to desire it in the first place.
This is like putting gold leaf on a turd. The companies in question treat users like datamines to be stripmined and left. These are listening posts that books like 1984 talked about. Yet unlike the stories of them being "placed", they're being sold to the gullible public who clamor to get the next version.
Echo&GHome < Project Alias < Not buying one at all
tldr; uses raspberry pi to continuously feed the smart speaker white noise until it receives your new keyword, "ok alias" or "ok computer" or whatever. I think it's cool, there have been countless times when I didn't even say "ok google" and it has randomly woken up
I get that this is an art project, but I think it would be more effective as a privacy device if it inserted a switch inline with the microphone, doing away with the "white noise" entirely. :)
"Project Alias offers an independent layer of protection to any privacy-minded person"
how>? that makes no sense. only because it wakes it for you?
i think its good that you need to talk to google or some company, at least then you have an honest picture where your data goes :D. most useless machine in the world, apart from the damn home-AI nonsense itself :D
I still think voting with your wallet and not buying the devices is the better tack. This just normalizes the notion that they are always listening, qnd the best we can do is pay for the hardware and then hack around it? Bollocks.
Well, props for trying to fix these assistants, but i'm not gonna pay more for what should be already. And, yes, I won't buy those things if I can't change the "name". It's like the first thing I want to do. (full disclosure: i do have a google mini that came as a promotional free gift)
- There is not evidence that these devices record and transmit without an activation word triggering this behavior - However, there is nothing to stop companies from breaking this assumption - Some people think the risk of one of these companies flipping a switch and recording everything is negligible - Some people think the risk of one of these companies flipping a switch and recording everything is warrants serious concern - These two groups will not agree, and that's fine :)