Quote:
Originally Posted by Uncle Robin
For those who are interested and/or concerned, and who have the inclination to do so, there's an interesting DIY detection exercise suggested in a followup article at El Reg |
That article suggests almost exactly the experiment jhowell suggested a few pages above.
What I found most interesting was the cited paper "When speakers are all ears: Characterizing misactivations of iot smart speakers."
It addresses and documents the unintentional activation of smart speakers, which can result in them sending audio to the cloud that the user did not intend, but what matters more to me was they proved with some certainty that these devices are NOT always sending audio home and do not appear to be behaving with any malicious intent or intentional subterfuge. They just have less than perfect wake word detection.
So I can continue to rest easy.