Slashdot Mirror
Samsung SmartTV Customers Warned Personal Conversations May Be Recorded
An anonymous reader writes Samsung's privacy policy includes details that its Smart TV voice recognition feature may pick up on personal conversations and transmit private communications to third parties. Buried in the privacy policy related to the smart television, Samsung advises users to be aware that any snippets of conversation might be captured by the software which allows them to control their television sets with a series of commands. Questions have been raised about who these third parties could be, what the information is used for, and how the data is being transmitted – with potentially unencrypted voice clips left exposed to hackers.
and I went out of my way to AVOID samsung. at costco, almost all the sets (in the store) were sammys. sucks! 2nd popular brand was vizio and I decided to give that one a try.
anything over 39" was ONLY avail in 'smart' tv format.
fortunately, I was not forced to accept an eula and I never enabled the smart mode. good, actually, since when you give it IP connectivity, it auto updates itself and the current version is bad (no one likes it).
I hope that by continuing to deny it access, it won't ever decide on its own to go snooping for open wireless APs. that would be really bad.
then again, at some point when the new firmware is worth getting, I'll have no choice but to enable IP ;( I don't think you can just carry the firmware over with usb; they don't give you firmware, they only update it 'live over the net'.
its sad that you can only buy smart tv's at a certain size or bigger. I expect the only a few really low end models will be non-smart, as time goes on. nothing I hate more than paying for stuff that I don't want and refuse to even enable.
my htpc does all the 'smarts' I want. my files have no drm and so I don't need or want anthing smarter than vlc on win7, for example.
samsung is bad news, though. pretty evil as a company. they have the rep of building things that last 'the warranty period + 1 day'. its almost literally true, too; they try to use parts that will last a very short time (eg, electrolytic caps). samsung has no ethics at all. its sad that they have so much market share in so many things these days.
--
"It is now safe to switch off your computer."
how is it remotely avoidable when your tv has voice recognition? like systems like siri it'll send it off to a remote server for analysis and to see what it has to do :). It will pick up random conversations and try to see if it's a command. I'm not endorsing things like this, i would never want a smart tv, i'll attach a htpc or raspberry pi or whatever to it if i want such features. But i can understand that things like that are pretty unavoidable with voice commands, and don't sound that much like a spying nightmare, just a logical result of voice recognition...
A TL;DR: Samsung TVs use voice commands and like most other voice command services they outsource voice recognition to their server park. Unlike other services, however, they continously listen and send this data, they don't wait for a button press or keyword.
I would like them to explain why a recording function is needed in the first place.
Probably for the voice recognition. The way smartphones have been doing it (and the way I'd suspect this TV works as well) is that they record your spoken commands, then send them out over the internet for a more powerful machine to crunch. After a few milliseconds of processing time, the interpretation is sent back to the phone and it performs the commands.
The reality is that this is probably a tempest in a teapot and samsung isn't doing anything more nefarious than apple does with siri or google does with the "ok google" feature on android phones. That said, samsung deserves all the flak they get over this. They should have known better than to leave that kind of blanket statement in their license agreement as it clearly allows for abuse on the part of samsung or their 3rd parties.
Competent natural-language voice recognition is still too hard for a handheld or embedded device.
I disagree.
First, doesn't the XBox One do it without internet access?
But in general, we had competent natural-language voice recognition in 2002 on single-core x86 CPUs. Today's embedded ARMs are just as capable as those. I personally worked on a project where the software could recognize phrases such as "Patient presents with acute myocardial infarction caused by necrosis of myocardial tissue..." with very good accuracy. I know this is harder to do without context, so that same program might not get "Okay Google, show me pictures of Tom Cruise in his new car." But it sure as heck should be able to get "Samsung Play movie" and "Samsung volume up" every time.
We even had "dumb" phones in the 1990s that could recognize "Call Kathy Smith" where it could recognize a name from your contact list. That's not so different from "Play Game of Thrones."
(I actually work on stuff about this, thus anon.)
The idea is that the recognition of the "ok, google" or "siri" activator is simple enough that it runs on the device, without needing to send it to the cloud. Once the magic word wakes the device up, it starts sending audio up to the ASR remote service.
You didn't press the voice-to-text microphone icon. You hit the "Capture a sound" microphone icon, which is a new feature in the iOS 8 Message apps (https://www.apple.com/ios/whats-new/messages/ scroll to "Add your voice to the conversation").
Now whether having two microphone icons with two different behaviors on a single screen is good UI design is whole other matter...