We are delighted to announce that Hong-Hanh has been awarded a competitive grant from the ENFIELD project (funded by Horizon Europe) to advance the state-of-the-art in voice security. The proposal, titled “Generalization of Deepfake Voice Detection across Generator Architectures,” has secured €24,000 in funding to address critical challenges in biometric security.
Project Overview. As generative AI evolves rapidly from Generative Adversarial Networks (GANs) to Diffusion Models (DMs), current deepfake audio detection systems often fail to generalize across these different architectures. To address this challenge, Hong-Hanh aims to develop a system that can generalize across different generator families (e.g., GANs and Diffusion Models).
International Collaboration. The project will start in January, with a 4-month exchange at Budapest University of Technology and Economics (BME) to foster international collaboration. Hong-Hanh will collaborate with Prof. Géza Németh at BME to test and refine these models on state-of-the-art datasets.
Impact. The project aims to deliver novel methods to verify the system’s strength against manipulated voice samples. This research strengthens Europe’s digital forensics capabilities, providing crucial tools for combating synthetic media threats.