Cyborgs & Biohacking

Identity thieves talk with manipulated original voice

Have others say what they never actually said – much less wanted to say. This is currently only possible via remix, but not for long: Voice conversion makes entirely new sentences out of voice recordings.

03 Mär. 2017
Identitätsdiebe Cyber Security
(Photo: YouTube/Adobe)

The "yes catchers" have now become established in Germany: callers from dubious companies that elicit a "yes" with questions such as "Can you hear me?", which they later reedit into a realistic-sounding order confirmation using audio editor. What follows are expensive bills that you must defend yourself against. The federation of German consumer organizations provides advice on how to avoid saying the word "yes" at all. Their website also includes sample letters and a link to the complaints office of the Federal Network Agency – because the method is, of course, criminal.

It will soon be easier for such con artists to obtain specific original sound words. Adobe has developed the equivalent of Photoshop: VoCo needs only 20 minutes of voice recording to generate any combinations on this basis (the acronym stands for voice conversion). You don’t even need experience as a sound engineer to do so, as the software converts the original recordings into text. The user only has to rewrite the wording – click! – and VoCo converts the new sentence back into an authentic-sounding audio file. Given security systems that use the human voice for authentication , the potential of this application is quite alarming. Adobe is nonetheless at least working on a method for detecting manipulated voice recordings. It is, however, not possible to detect the fake with the naked ear.

Cyborgs & Biohacking Media & AdTechBiometricsCognitive ComputingSecurity CEBIT RSS Feed