Liveness detection challenge to prevent AI voice fraud launched by US regulator

The Federal Trade Commission has opened a $25,000 challenge looking for the best and broadest answer to AI-enabled voice cloning fraud.

Commissioners want to see biometrics entries that package products, procedures and policies. Submissions can be made here January 2-12. Rules are here.

James Evans, Christine Barker and Amritha Jayanti are the commission’s point people.

Winning entries have to demonstrate at least two of three capabilities. The most important is prevention or authentication. The second is real-time detection or monitoring of voice cloning use. And the last is post-use evaluation of suspected cloning.

Real-time detection of spoofed voices is known within biometrics as presentation attack detection, or voice liveness detection. Previous challenges for this type of technology include the SASV (spoofing-aware speaker verification) Challenge 2022, which was won by ID R&D.

The FTC also want to see entries that are easily managed and that put the bulk of responsibility and liability on vendors rather than consumers.

The top cash prize is followed by a $4,000 runner-up award and as many as three $2,000 honorable mentions. Submissions from groups of 10 or more can vie for the non-cash recognition award.

Senators are worried about the security of voice biometrics being used by banks, particularly in light of the increasing availability of voice-cloning tools.

Article Topics

biometric liveness detection  |  deepfake detection  |  fraud prevention  |  FTC  |  presentation attack detection  |  spoof detection  |  voice biometrics