Your Smart Speaker is Accidentally Activating Up to 19 Times a Day
- University researchers found out that smart speakers engage in false triggering a lot more often than we thought.
- That is especially the case when the user is watching a movie or a TV show where dialogues are involved.
- The worst of all was Amazon’s Echo Dot series 2, but the rest of the bunch didn’t do much better.
According to a recent study by researchers at Northeastern University and the Imperial College in London, smart speaker devices are activating by mistake by up to 19 times a day. The study tested out popular products that deploy voice assistants like Amazon’s Alexa, Google’s Assistant, Microsoft’s Cortana, and Apple’s Siri. Having too many false triggerings on these devices raises privacy concerns, as IoT (Internet of Things) devices are everywhere nowadays. This is especially worrying when many of the recordings that shouldn’t have been taken end up in the hands of human reviewers.
The researchers devised a refreshing testing approach that involved the playback of TV shows where a wide spectrum of different dialogues are heard. So, they fired up Netflix and played 125 hours of comedies, dramas, realities, mysteries, and crime series. To reach to safe conclusions, they repeated this test for twelve times. Nobody wants their IoT to activate while watching Netflix, so this is considered false triggering. In the same time, they set up a network traffic monitoring tool to detect when a smart speaker activates, and so here is a summary of their findings:
- None of the devices engaged in continuous and constant recording.
- The rate of false activations ranges between 1.5 and 19 times a day.
- The most prone to activation was the Apple HomePod and Microsoft’s Cortana. Echo Dot series 2, series 3, and Home Mini are also pretty inaccurate.
- The false activation occurs inconsistently, with only 8.44% of activations being repeated across all 12 tests.
- Shows with more dialogue like “The Office” and “Gilmore Girls” triggered the smart speakers more often.
- While the activations remained limited to short intervals, some lasted for 43 seconds, which is enough to capture sensitive data. Echo Dot 2nd Gen is the worst, with the recordings usually going beyond 20 seconds.
- The words that cause false activations the most are “Okay”, They”, “Hey”, “I’m”, and “Hi”. Of course, each AI falls for a different set of words, like Cortana, for example, which wakes up with the word “Colorado”.
If there’s a smart speaker around, you never know when it will start listening, so the best way to treat these devices is as if they are active the whole time. If you need privacy, pull the plug. If you find using smart speakers a convenience that you can’t live without, you’d better turn them on only during the times that you need them. Other than that, there’s not much else you can do.










