NAUSS-Spoof50: Arabic Speech Benchmark for Spoofing Detection

NAUSS-Spoof50 is a multi-task Arabic speech benchmark designed for audio spoofing detection and forgery localization. The dataset's metadata is freely accessible, while audio files are available upon request. It was created by Mostafa98 and released under a Creative Commons Attribution Non-Commercial 4.0 license.

Use Cases

• Detect fully synthetic speech spoofing based on the benchmark's multi-task design.
• Localize forgery in partially manipulated audio clips based on the dataset's description.
• Benchmark model performance for Arabic speech security tasks.
• Train machine learning models for audio authenticity verification.

Strengths

• Designed for multi-task evaluation, supporting both detection and localization.
• Focuses on Arabic speech, a specific and less common domain in spoofing benchmarks.
• Metadata is freely accessible for initial analysis.

Limitations

• Audio files are not freely downloadable; they are available only upon request.
• Column-level documentation and sample data are unavailable.
• Row count and file formats are unknown, limiting suitability assessment.

Provenance

Source

huggingface

Freshness

Last updated 2026-06-06 20:15:14.