XLSR-Mamba: A Novel Architecture for Efficient and Effective Spoofing Attack Detection in Speech Signals

This research paper introduces XLSR-Mamba, a novel deep learning model that combines a pre-trained XLSR model with a dual-column bidirectional Mamba architecture for superior performance in detecting spoofed speech, outperforming existing models in efficiency and accuracy, particularly on challenging datasets like In-the-Wild.