Alexa, how often do you listen in?
Ever wondered how many times your smart speaker accidently activates in a day? Researchers from Northeastern University did too, and here’s what they found…
Voice assistants like Amazon Alexa, Google Assistant, Apple’s Siri and Microsoft’s Cortana accidentally wake up and start recording what they hear up to 19 times a day.
Anyone who’s used a digital assistant knows they have a habit of ‘waking up’ – making a sound or lighting up to indicate they’re listening – when they shouldn’t. The systems are listening out for key phrases, or ‘wake words’ like “Hey Google” or “Siri”, before they turn on and start recording in anticipation of a voice query. But these digital eager beavers often jump the gun mistaking the word “seriously” for “Siri”, for example, as an invitation to listen in.
The Mon(IoT)r research group at Northeastern University in Boston, USA, are dedicated to investigating what privacy risks might arise from inviting more and more connected devices into our homes. They wondered how many times a day these misfires happened. To find out, they put a bunch of smart home speakers in a room together to watch 125 hours of dialogue-heavy Netflix content, including Gilmore Girls, Grey’s Anatomy and the West Wing. Some people get all the good jobs.
They discovered that the average rate of activations per device was between 1.5 and 19 times per day. No smart speaker was found to be consistently recording conversations, but they did wake up for short intervals at inconsistent times. In other words, when the experiment was repeated (12 times) the same phrases didn’t always wake up the speaker. Once on, the devices would typically record at least the next six seconds of conversation, though some assistants record more.
So the good news is that the assistants seem to be making honest mistakes, but the bad news is that in some cases they can be recording up to 19 times per day, without meaning to. In further research the team wants to investigate how much of this data is being sent to the cloud, whether these accidental recordings are reported accurately, and whether there are any other factors at play.
“ONCE ON, THE DEVICES WOULD TYPICALLY RECORD AT LEAST THE NEXT SIX SECONDS OF CONVERSATION”