"How Voice Assistants Follow Inaudible Commands"

As the popularity of voice assistants continues to rise, concerns surrounding the security of speech recognition systems grow. Security researchers have discovered that it is possible for attackers to send malicious inaudible voice commands to speech recognition systems over the air. Thorsten Eisenhofer, an IT security researcher at Ruhr University of Bochum, supports the use of the MP3 principle as a countermeasure against such attacks. According to Eisenhofer, speech recognition systems should be combined with an MP3 encoder to delete any ranges that are inaudible to the human ear before they reach the systems. This method would not stop attackers from manipulating audio files, but it would prevent their attacks from being hidden. This article continues to discuss the attacks that have been demonstrated on speech recognition systems and how the security of these systems can be strengthened by using the MP3 principle. 

RUB reports "How Voice Assistants Follow Inaudible Commands"