Resumen de A speech jammer using ultrasonics for speech privacy protection and smart speaker attack prevention
Kazuhiro Kondo, Noriyuki Hayasaka
We investigated a speech-jamming device that uses modulated ultrasonic signals to jam hidden recording devices, as well as to avoid stealth injection of speech commands to smart speakers. Recordings are jammed using demodulated audible noise signals due to the non-linearities in analog circuits in the recording devices. Additionally, this ultrasonic signal should not be audible to humans to avoid unnecessary annoyance. We find that the effective modulation schemes for jamming seem to be highly device dependent, but frequency-modulated signals seem to work modestly well with the devices tested. We then created a prototype jamming device that creates a modulated ultrasonic jamming signal uniformly in the surrounding environment. The jamming signals were simultaneously output from the ultrasonic transducers placed on the dome-shaped shell of the prototype. We evaluated the jamming capability of this prototype by recording speech samples played out from a loudspeaker when jamming signals are simultaneously output from the prototype. It was confirmed that the prototype effectively lowers the intelligibility of the recorded speech samples uniformly in most of the surrounding area.
