In a nutshell: Fraunhofer is developing person-centric voice profiles to detect audio deepfakes as general detection methods can no longer keep pace with rising synthesis quality.
Classical detection methods for AI-generated voices are reaching their limits. The Fraunhofer Institute’s PADSE research project is therefore developing a person-centric approach that should identify audio deepfakes via individual speech and voice profiles.
The Fraunhofer Institute for Digital Media Technology (IDMT) is working on a new method for detecting audio deepfakes. The approach is not based on analyzing universal features of manipulated speech, but rather on capturing and comparing individual voice profiles.
For CISOs, this represents a necessary expansion of previous security strategies: as the quality of speech synthesis increases, attacks via forged audio recordings – for example to manipulate employees or for authentication purposes – are becoming increasingly plausible. Traditional, general detection patterns no longer suffice for these applications.
PADSE’s person-centric approach promises higher accuracy by capturing typical characteristics of a known voice – similar to a fingerprint. Deviations in speech rate, pitch or articulation can then be specifically detected. This enables organizations to protect identity verification and communication channels more robustly at the audio level.
Source: www.security-insider.de · Published 20 July 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.