DSRI Sponsors SAFE Audio Forensics Challenge to Identify Synthetic Audio 

The Digital Safety Research Institute (DSRI) sponsored the SAFE: Synthetic Audio Forensics Evaluation Challenge, a global research competition focused on detecting synthetic and manipulated audio. Hosted on the Hugging Face platform and culminating at the IH&MMSEC 2025 workshop on June 18-20.  The challenge brought together AI researchers from around the world to tackle identification of machine-generated and manipulated audio. 

Participants were challenged with three distinct detection tasks, each requiring analysis of audio samples manipulated through techniques such as generation, compression, and laundering. Submissions were evaluated in on blind datasets to ensure comparative results. The overarching goal was to advance the accuracy of audio forensic tools. 

We’re proud to announce that the Image and Sound Processing Lab at the Polytechnic University of Milan has emerged as the top performer in the SAFE Challenge. 

In recognition of the outstanding contribution, DSRI will be offering the Polimi team a $500,000 research grant over two years to further develop its audio forensic analytic. The grant would enable the team to research explainability of their analytic to help users understand the rationale behind the analytic result. This initiative highlights DSRI’s commitment to advancing digital safety by investing in research-driven solutions that address today’s—and tomorrow’s—emerging threats. 

Learn more about the SAFE Challenge and future updates at: https://stresearch.github.io/SAFE/