Towards Explainable Spoofed Speech Attribution and Detection: a Probabilistic Approach for Characterizing Speech Synthesizer Components Journal: Computer Speech And Language

Jagabandhu Mishra, Manasi Chhibber, Hye-jin Shim, Tomi H. Kinnunen · 2025

Read Towards Explainable Spoofed Speech Attribution and Detection: a Probabilistic Approach for Characterizing Speech Synthesizer Components Journal: Computer Speech And Language by Jagabandhu Mishra, Manasi Chhibber, Hye-jin Shim, Tomi H. Kinnunen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-04049.txt.

Description

Jagabandhu Mishra and co-authors build a 2025 method to both spot fake speech and name the tool that made it. Fake speech from synthesizers can fool listeners and systems. Most detectors give a bare yes-or-no answer from dense internal features that no one can read, which makes their choices hard to trust or explain.

Their method breaks a voice clip into a set of named traits, each with a chance value, that point to parts of the synthesizer that made it. Those trait scores feed simple classifiers that handle both tasks: telling real from fake, and naming the source method. They also use a game-theory tool to weigh how much each trait swayed the final call. The result is a detector whose reasoning can be inspected.

Catalog record

CREATORJagabandhu Mishra, Manasi Chhibber, Hye-jin Shim, Tomi H. Kinnunen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE99 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-04049

Browse the full library · About the Tower of Babel