Synthetic voice forgery in the forensic context: a short tutorial

Guillaume Galou (1) , Gérard Chollet (1)

1 LTCI - Laboratoire Traitement et Communication de l'Information

Technical voice forgery in the forensic area has led to several studies, mainly dealing with voice conversion. In the last decade, latests developments around voice synthesis have reached satisfactory intelligibility and quality levels. Moreover, several web-based or standalone apps can be used for TTS. Nowadays, HMM-based synthetic voices can be built to fool biometric systems. Several authors reported FARs as high as 70 to 80% when TTS voices where used. Nevertheless, the forensic context is quite different since the human ear might be able to detect a synthetic voice, thus leading to a case discarding. We used the MARY TTS platform in order to produce a speaker-dependent synthetic target voice sample. Given a single synthetic trial, our very preliminary work is to show how feasible and successful such an attack is. Further work is needed to build other voices and provide accurate statistics. Our aim is to confirm a criminal opportunity and to show that synthetic voice detection might become mandatory in a near future.

Mots clés: TTS; speech synthesis; forensic; speaker recognition; voice comparison; HMM; HTS; MARY Text-to-Speech; voice forgery; voice disguise

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Guillaume Galou, Gérard Chollet. Synthetic voice forgery in the forensic context: a short tutorial. Forensic Speech and Audio Analysis Working Group (ENFSI-FSAAWG), 2011, Rome, Italy. ⟨hal-00625918⟩