All those “intelligent” voice assistants – Google, Siri, Alexa, smart cars (not not “Smart cars”), etc. all have a seriously damaging misfeature. It is most evident (and indeed amusing) when you are watching a review of a smart car.
All of these voice assistants activate when they hear a particular sound pattern (a word) which is fixed. Now there are probably technical reasons why it is fixed, but that does not mean they cannot be overcome.
I don’t want to address a smart device by a name that I cannot change. I want to shout “Slave! Get me a cup of coffee”, not “Alexa! Phone Alex”. Which is of course a nice to have feature, not a bug.
But what about about attacks? Say “Okay Google, turn off all alarms” (and your victim is late to work), say “Okay Google, set an alarm for 3am” (and your victim is grumpy at work).
That ‘trigger word’ (or phrase) is in effect a password – not a very secure one, but one never-the-less.
Lastly there is what hilariously happens when you see a review of a new car with a voice assistant made by a certain car company whose name starts with “M”. Because of course a reviewer will use the “M” word whilst describing the vehicle, and the miscommunication with the assistant is a hilarious indication of how rubbish voice assistants are (they aren’t really but that’s the impression you get).
Make the trigger word/phrase adjustable, and this all goes away.
