Cross-Domain Sequential Recommendation (CDSR) leverages diverse behavioral data across domains to alleviate the data sparsity issue commonly encountered in single-domain scenarios. However, the potential of intra-domain information remains underexplored in existing CDSR models, despite its importance in modeling domain-specific user preferences. Furthermore, they typically capture the users’ preferences independently of each domain, which may neglect the cross-domain sequential dependencies from different domains. To alleviate these issues, we propose Intra-domain preference modeling with Cross-domain dependency alignment for Cross-Domain Sequential Recommendation(IC-CDSR), which models intra-domain preferences and aligns cross-domain dependencies. Specifically, IC-CDSR consists of two main modules: a cross-domain collaborative modeling module (CDCM) and a cross-domain information alignment module (CDIA). The CDCM module comprises two key components: a selective GRU, which integrates temporal convolution and gating mechanisms to extract fine-grained contextual dependencies and capture users’ local preferences within each domain; and a user-anchored cross-attention encoder, which leverages a time-invariant user embedding to attend over behavioral sequences, thereby capturing global preferences and enabling effective cross-domain integration. The CDIA module employs a cross-attention encoder with Rotary Positional Encoding(RoPE), which encodes relative positional information through rotation within the attention mechanism, thereby preserving semantic integrity while aligning cross-domain behaviors and capturing sequential dependencies between source and target domains. Experimental results on the Amazon Book, Movie, and CD dataset validate the effectiveness of IC-CDSR against state-of-the-art(SOTA) baselines.