Idiap’s postdoc receives BRIDGE Proof of Concept grant for audio deepfake detector

Ajinkya Kulkarni, a postdoctoral researcher at Idiap, has been awarded an SNSF Bridge Proof of Concept grant. His project addresses one of the fastest-growing threats in digital security: voice fraud. Audio deepfakes, AI-generated recordings that convincingly mimic real voices, are already being used for identity fraud, investment scams, and political disinformation.

Ajinkya’s project, called Proof-of-Voice (PoV), aims to develop an audio deepfake detector capable of running directly on everyday devices such as smartphones, browsers and laptops. The goal is to enable individuals, SMEs and online communities to verify suspicious audio in real time without sending sensitive biometric data to external servers.

The project builds on earlier research that detects fake voices, aiming to turn it into an easy-to-use tool and test it in real-world situations so devices can spot audio deepfakes directly.

Throughout these initial deployments, the project will collect metrics to assess both the technical robustness and the commercial viability of the solution across different environments.

The expected outcome is a validated, privacy-preserving on-device detection system and a ready-to-integrate toolbox for industrial applications, showcasing real-world impact.

When talking to him about the project, Ajinkya explains the reason behind his project’s idea: “Where there is AI content so are the deepfake and thus the need for detectors”.

Beyond deepfake detection, the project lays the foundation for a broader class of lightweight speech models capable of supporting tasks such as speaker verification, authenticity checks and secure human–AI voice interaction. In doing so, it positions Swiss research and industry as key contributors to the development of trustworthy and secure speech AI technologies.

We look forward to following the progress of this project!