NAUSS-Spoof50: A Multi-Task Arabic Speech Benchmark for Spoofing Detection

NAUSS-Spoof50 is a curated multi-task Arabic speech benchmark designed to support both fully synthetic spoofing detection and partially manipulated forgery localization. The dataset's metadata is freely accessible, while audio files are available upon request. It was created by Mostafa-Moallim and last updated on June 7, 2026.

Use Cases

• Train models for fully synthetic audio spoofing detection based on the benchmark's multi-task design.
• Develop algorithms for localizing partially manipulated forgeries within speech audio.
• Benchmark performance of speech forensics systems on Arabic language data.
• Research the characteristics of spoofed and genuine Arabic speech signals.

Strengths

• Designed as a multi-task benchmark supporting both detection and localization tasks.
• Specifically focuses on Arabic speech, addressing a particular linguistic domain.
• Metadata is documented and freely accessible for evaluation.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Audio files require a separate request and are not immediately downloadable.

Provenance

Source

huggingface

Collection Method

Curated benchmark, likely assembled for research purposes.

Time Range

null

Freshness

Last updated 2026-06-07 07:12:45; freshness should be verified.

Geography

Focus on Arabic speech, but specific geographic coverage is not detailed.