FoeGlass uses in-context learning for red teaming audio deepfake detectors

Arxiv CS.SD (Sound/Audio ML)

Paper proposes FoeGlass, a simple in-context learning method for red teaming audio deepfake detectors. It generates test samples to identify weaknesses in state-of-the-art ADD models. The approach requires no additional training and can be applied to any TTS model.

in-context-learningred-teaming

Arxiv CS.SD (Sound/Audio ML)·Sepehr DehdashtianJacob H SeidmanVishnu N BoddetiGaurav Bharaj

7 days ago