2026-10-06T07:27:55Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34018134 2026-09-28T18:12:22Z category_26407 item_type_12 month_year_09_2026

Same Bits, Different Verdict: The Decoder in Speech Deepfake Detection GANG SHI (25132929) Signal processing speech deepfake detection audio deepfake detection anti-spoofing speech coding Opus decoder-side processing receiver-side rendering threshold calibration domain shift cover-source mismatch <p dir="ltr">Speech deepfake detectors consume decoded waveforms, but for coded speech the bitstream is what was transmitted; which rendering of it the detector sees belongs to the receiver. Decoding a bit-identical Opus bitstream with libopus 1.6.1’s stronger optional decoder-side post-filter enabled rather than disabled shifts class-mean detector scores toward the synthetic class in 20 of 20 detector–class–corpus cells, with no packet loss and no change to a transmitted byte. At a threshold transferred from the unenhanced rendering at a 1% false-alarm target, false alarms rise in all eight cells; one public detector goes from 1.00% to 3.80% while its equal-error rate moves only from 0.30% to 0.40%. The shift is not a monotone rescaling that recalibration absorbs: items cross the threshold in both directions, and equal-error rate moves in every cell. On held-out data a stale threshold multiplies false alarms by 1.7 to 3.8. Refitting recovers the target rate, but no corpus, protocol or file header records that the configuration changed.</p><p dir="ltr"><br>This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible.</p> 2026-09-28T18:12:22Z Text Preprint 10.6084/m9.figshare.34018134.v1 https://figshare.com/articles/preprint/Same_Bits_Different_Verdict_The_Decoder_in_Speech_Deepfake_Detection/34018134 CC BY 4.0