|
|
|
|
|
by dantle
2423 days ago
|
|
I'm a bit surprised that this works. It appears that this attack targets a single microphone. However, internal to most of these home assistant devices (e.g. Echo or HomePod) there is an array of microphones. A real sound from a spoken word would probably show up on more than one microphone (with an appropriate time and phase offset), although it seems to not be currently required. It seems like it would be complex or impossible for an attacker to target more than one microphone with this attack. |
|