<p>Behavioral biometric recognition on Wi-Fi struggles with generalization and robustness. To tackle these, Wi-IDSync, a multi-domain behavioral biometric recognition framework, is proposed. In this framework, the Mamba block is adopted to process discrete channel state information, capturing fine-grained behavior features. The shared class-level information is captured through stable prototypes and pseudo-clustering in prototype-based contrastive learning, achieving class-level domain alignment. The proposed CSI-oriented multi-layer domain alignment module, MA-DAM, integrates attention-based fusion with adversarial alignment to mitigate multi-state domain shifts. The fused feature is mapped through a shared bottleneck, and a consistency regularizer is adopted to stabilize training. The channel swapping module and multiple kernel maximum mean discrepancy function further optimize alignment of data distributions across the source and target domains. The effectiveness of Wi-IDSync is verified through data of volunteers with multiple popular personal states. Experimental results show that Wi-IDSync achieves outstanding behavioral biometric recognition accuracy in both single-domain and multi-domain scenarios, demonstrating its satisfactory performance.</p>