# From Scores to Evidence: Auditable Decisions Can Improve Speech Deepfake Detection

arXiv: https://arxiv.org/abs/2609.08899 · HTML: https://mlsanity.com/paper/2609.08899/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 28 of its 31 references that are in the map.
- First posted on arXiv 8 Sep 2026.

## Builds on (most cited first, 28 of 28)

- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://mlsanity.com/paper/2106.07447/) (2021)
- [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://mlsanity.com/paper/2110.13900/) (2021)
- [Selective Classification for Deep Neural Networks](https://mlsanity.com/paper/1705.08500/) (2017)
- [AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks](https://mlsanity.com/paper/2110.01200/) (2021)
- [End-to-end anti-spoofing with RawNet2](https://mlsanity.com/paper/2011.01108/) (2020)
- [ASVspoof 5: Crowdsourced Speech Data, Deepfakes, and Adversarial Attacks at Scale](https://mlsanity.com/paper/2408.08739/) (2024)
- [WaveFake: A Data Set to Facilitate Audio Deepfake Detection](https://mlsanity.com/paper/2111.02813/) (2021)
- [Proactive Detection of Voice Cloning with Localized Watermarking](https://mlsanity.com/paper/2401.17264/) (2024)
- [ASVspoof 5: Design, Collection and Validation of Resources for Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech](https://mlsanity.com/paper/2502.08857/) (2025)
- [Post-training for Deepfake Speech Detection](https://mlsanity.com/paper/2506.21090/) (2025)
- [From Sharpness to Better Generalization for Speech Deepfake Detection](https://mlsanity.com/paper/2506.11532/) (2025)
- [ASVspoof 5: Evaluation of Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech](https://mlsanity.com/paper/2601.03944/) (2026)
- [Revisiting and Improving Scoring Fusion for Spoofing-aware Speaker Verification Using Compositional Data Analysis](https://mlsanity.com/paper/2406.10836/) (2024)
- [Self Voice Conversion as an Attack against Neural Audio Watermarking](https://mlsanity.com/paper/2601.20432/) (2026)
- [Spoof Diarization: "What Spoofed When" in Partially Spoofed Audio](https://mlsanity.com/paper/2406.07816/) (2024)
- [Spoofing-Aware Speaker Verification Robust Against Domain and Channel Mismatches](https://mlsanity.com/paper/2409.06327/) (2024)
- [A Comparative Study on Proactive and Passive Detection of Deepfake Speech](https://mlsanity.com/paper/2506.14398/) (2025)
- [Explaining Speaker and Spoof Embeddings via Probing](https://mlsanity.com/paper/2412.18191/) (2024)
- [LENS-DF: Deepfake Detection and Temporal Localization for Long-Form Noisy Speech](https://mlsanity.com/paper/2507.16220/) (2025)
- [Towards Data Drift Monitoring for Speech Deepfake Detection in the context of MLOps](https://mlsanity.com/paper/2509.10086/) (2025)
- [Zero-Day Audio DeepFake Detection via Retrieval Augmentation and Profile Matching](https://mlsanity.com/paper/2509.21728/) (2025)
- [FakeMark: Deepfake Speech Attribution With Watermarked Artifacts](https://mlsanity.com/paper/2510.12042/) (2025)
- [Deepfake Word Detection by Next-token Prediction using Fine-tuned Whisper](https://mlsanity.com/paper/2602.22658/) (2026)
- [A Preliminary Case Study on Long-Form In-the-Wild Audio Spoofing Detection](https://mlsanity.com/paper/2408.14066/) (2024)
- [The Watermark Shortcut: How Provenance Marking Sabotages Audio Deepfake Detection](https://mlsanity.com/paper/2606.23335/) (2026)
- [Toward Interpretable Speech Deepfake Detection using Artifact-Specific Experts and Calibrated Detection Scores](https://mlsanity.com/paper/2607.21127/) (2026)

## Abstract (arXiv)

Speech deepfake detectors usually emit one score per utterance, but a borderline score does not reveal why two examples differ during retrospective error analysis. We ask whether a final score can be calibrated from component fields while keeping those fields visible for inspection. We build a decision record with a passive detector score and a score from a probe applied to a marked copy. It also includes retrieval support held out of the evaluated family, a margin from a support-set profile, and raw neighbor closeness. A cross-fit calibrator combines these fields and two differences between raw scores into one final score. On matched ASVspoof development data, the calibrated record reduces equal error rate (EER) by 3.48 percentage points relative to the fixed retrieval-augmented rule. It reaches 8.43% EER, whereas a passive WavLM baseline reaches 6.71% on the same subset. The record is therefore not the strongest detector in this comparison. Its value is to retain inspectable component fields while producing one scalar score for retrospective diagnosis.