I Can Hear You Now: The Growing Danger of Voice Recognition
Privacy Plus+
Privacy, Technology and Perspective
I Can Hear You Now: The Growing Danger of Voice Recognition. We have written about the dangers of facial recognition and have called for a moratorium while the privacy issues it poses are explored and discussed.
Voice recognition is just as dangerous. The Wall Street Journal published an article in August that “deepfake” technology was responsible for a voice phone call (vphishing) that impersonated the CEO of a company and fraudulently convinced another person within the company to transfer over $200,000. A link to that article follows:
It seems clear then that the automation of cyberattacks is upon us. More than that, however, lies are now being disguised to sound like truth.
For the spoken word, however, a sense of sanctity is at least as old as the human voice, and has been protected against eavesdropping since Sir William Blackstone’s day. According to Blackstone, in his Commentaries on the Laws of England (1769): “eavesdroppers, or such as listen under walls or windows, or the eaves of a house, to hearken after discourse, and thereupon to frame slanderous and mischievous tales, are a common nuisance and presentable at the court-leet.”
As technology has evolved, intercepting and divulging the spoken voice, and, in particular, telephone calls (at least without consent or permitted exceptions) has been unlawful under federal law since at least 1934; and since then, restrictions have been expanded into the digital age. States have been able to overlay their own restrictions, for example by requiring all-party consent before a call may be recorded. Some state statutes specifically prohibit interception of “aural” communications while leaving video recordings open – which is one reason why most CCTV systems do not activate whatever aural recording features they have.
But a new day is dawning. With neural data networks and artificial intelligence, eavesdropping is now not just about what we overhear, it is capturing and exploiting the voice itself.
The ability to recognize and authenticate a person by her voiceprint alone has become sufficiently stable and reliable that financial institutions and their affiliates are now allowed to use it under 15 U.S.C. § 689. Google’s Android Q (which Google says will run on a user’s device), Apple Siri, Amazon’s Alexa, Microsoft’s Cortana, and others not only recognize and authenticate their “owners’” voices, but interpret their instructions and execute commands (adding – and perhaps correlating – this personal data with the databases they are collecting from other sources). And now, technology will do more than just authenticate a person – it can identify a person, and single that person out of a crowd of voices.
Just like facial recognition.
Worst of all, the same types of software and related databases make it possible for bad actors to mix and alter syllables, inflections, and patterns from earlier recordings, and to generate entirely new – and wholly bogus – “recordings” of a person “saying” things that they did not say. Combine faux recordings with faux video, and “deepfake” technology will leave us unable to trust either our eyes or our ears.
The implications for our democracy are obvious. Like facial recognition technology, this privacy issue needs careful analysis and a democratic discussion – now.
Hosch & Morris, PLLC is a Dallas-based boutique law firm dedicated to data protection, privacy, the Internet and technology. Open the Future℠.