LJ-Spoof: A Generatively Varied Corpus for Audio Anti-Spoofing and Synthesis Source Tracing

2026-01-12

AI Summary

This work addresses the limitations of current audio anti-spoofing research, which suffers from a lack of high-quality, speaker-specific data with systematic variations, hindering robust spoof detection and fine-grained attribution of synthetic origins. To bridge this gap, we present the first large-scale, high-variability speaker-specific corpus, encompassing 30 text-to-speech (TTS) model families, 500 generative variant subsets, and over three million utterances. The dataset systematically manipulates key factors—including prosody, vocoders, generation hyperparameters, source real speech, training strategies, and neural post-processing—enabling controlled analysis of their impact on spoofing artifacts. Designed to serve both as a training resource and an evaluation benchmark, the corpus significantly enhances anti-spoofing performance and source attribution accuracy under speaker-conditioned settings, thereby filling a critical void in high-quality data for the field.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Large Multimodal Models (LMMs)
• Application Domains: Humanities & Computational Social Science

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Security and Privacy: Large-scale security measurements

Abstract

Speaker-specific anti-spoofing and synthesis-source tracing are central challenges in audio anti-spoofing. Progress has been hampered by the lack of datasets that systematically vary model architectures, synthesis pipelines, and generative parameters. To address this gap, we introduce LJ-Spoof, a speaker-specific, generatively diverse corpus that systematically varies prosody, vocoders, generative hyperparameters, bona fide prompt sources, training regimes, and neural post-processing. The corpus spans one speakers-including studio-quality recordings-30 TTS families, 500 generatively variant subsets, 10 bona fide neural-processing variants, and more than 3 million utterances. This variation-dense design enables robust speaker-conditioned anti-spoofing and fine-grained synthesis-source tracing. We further position this dataset as both a practical reference training resource and a benchmark evaluation suite for anti-spoofing and source tracing.

Problem

• audio anti-spoofing
• speaker-specific
• synthesis-source tracing
• generative variation
• dataset

Innovation

• generative diversity
• speaker-specific anti-spoofing
• synthesis-source tracing
• systematic variation
• TTS corpus