Big thinkers to prototype deepfake detection tools in year-long challenge

Germany’s Federal Agency for Jump Innovations (Sprind) has issued a call to action for innovative thinkers to develop “breakthrough innovations to identify deepakes.” Citing the rapid development of AI deepfake capabilities and its attendant “fascinating and disturbing possibilities” – emergent as the so-called deepfake ecosystem – the agency is hosting a deepfake Funke, which translates as “spark,” to prototype deepfake detection and prevention tools.

Sprind’s website says the aim is to develop a comprehensive, scalable prototype that includes “reactive measures for detection and/or preventive measures to protect against image deepfakes and can be integrated into existing digital infrastructures.” Prototypes must be able to reliably detect and authenticate deepfakes across at least three different use cases.

The twelve selected teams each receive up to 350,000 euros for the first stage of the Funke, which lasts until May 2025. An evaluation will then determine which teams show “the greatest potential for innovation.” Those selected will move on to the next stage, and receive an additional €375,000 each in funding until the project concludes in November.

The project was launched on behalf of the Federal Ministry for Digital and Transport (BMDV). Teams selected for stage one come from research institutions, universities and startups in Germany, France and Slovenia. Their approaches span multimodal deepfake detection, analysis of pixels and image noise, watermarking and blockchain technology, among other deepfake detection and prevention techniques, reflecting Sprind’s goal to cover the “entire value chain of digital media processing.”

Sprind’s website notes that, “as the performance of systems and language models based on artificial intelligence (AI) and machine learning (ML) increases, not only the quantity but also the quality of deepfakes is improving rapidly. This poses risks – for trust, security and the perception of reality in our society.” The solution is clear, although the challenges are significant.

“We can only make progress if we use a fast, reliable and adaptable system to detect and prevent deepfakes,” says a man in a promo video for the project – pointedly not confirmed as a real human, or just another deepfake in a swelling sea of millions.

Article Topics

deepfake detection  |  deepfakes  |  fraud prevention  |  Germany  |  research and development  |  SPRIND