Descriptor: Bangla Audio Spoofing With Four Conditions and Real Speech (BAS4R)

This paper introduces BAS4R, a large-scale, multi-condition Bangla speech dataset designed for gender-aware real and spoofed voice analysis under diverse acoustic and signal manipulation scenarios. Although Bangla is one of the most widely spoken languages worldwide, publicly available datasets supporting robust voice authenticity verification and spoofing detection research remain limited. To address this gap, BAS4R comprises speech recordings collected from 200 native Bangla speakers, including 110 male and 90 female participants, representing diverse age groups. The inclusion of multiple age groups captures age-related acoustic variability, including pitch variation, physiological changes in vocal fold characteristics, voice strength differences, and spectral diversity, thereby reflecting natural vocal variation across the population. The dataset is organized into four structured classes-Female Real, Female Spoofed, Male Real, and Male Spoofed-captured under multiple real-world degradation and replay-based manipulation conditions. In total, the corpus contains approximately 143.88 hours of audio, including 5.75 hours of clean authentic speech, while the remaining data consist of systematically generated spoofed samples. Spoofing conditions include channel-based distortions, such as telephone and VoIP transmission, replay-based attacks, signal-processing transformations, including pitch shift, formant shift, and time stretch, and effect-based modifications, such as robotic and monster-like voice alterations. All audio files are systematically organized and consistently labeled to support reproducible experimentation. By relying exclusively on physical replay devices and real communication channels without synthetic text-to-speech augmentation, BAS4R preserves authentic device- and channel-induced acoustic artifacts. The dataset supports research in voice anti-spoofing, spoofing detection, gender-aware classification, and robust speaker verification, and aims to serve as a foundational resource for secure and trustworthy Bangla speech processing research.