What Counts as Real? Speech Restoration and Voice Quality Conversion Pose New Challenges to Deepfake Detection

Researchers demonstrate that current audio deepfake detection systems incorrectly classify legitimate speech processing technologies like voice conversion and restoration as fake audio. A new multi-class detection approach shows improved accuracy by distinguishing between authentic speech, benign modifications, and actual spoofing attempts.

• →Traditional binary deepfake detection systems misclassify legitimate speech restoration and voice conversion as spoofed audio.
• →Benign audio transformations create distributional shifts in AI models that reduce their ability to distinguish real from fake speech.
• →A multi-class classification approach better preserves spoof detection while reducing false positives from legitimate audio processing.
• →Current anti-spoofing systems model raw speech distribution rather than actual speaker authenticity.
• →The research highlights fundamental flaws in how AI systems determine audio authenticity in real-world applications.