Sensors-Driven Multimodal Deepfake Detection: A Cross-Attention Fusion Approach with Adaptive Modality Gating | Synapse