2026
SONAR: Spectral‑Contrastive Audio Residuals for Generalizable Deepfake Detection
ICML 2026poster
Deepfake audio detectors often fail to generalize to unseen attacks, in part due to \emph{spectral bias}: neural networks prioritize low-frequency structure while under-exploiting subtle high-frequency (HF) artifacts left by generative models. We introduce \textbf{SONAR} (Spectral-cONtrastive Audio …