<p>Cyberspace Surveying and Mapping (CSM) involves the identification and analysis of digital assets to support network management and security, yet its domain-specific named entity recognition (NER) remains underexplored. A key challenge is the semantic gap between general-domain corpora and CSM domain texts, the suboptimal performance of existing named entity recognition (NER) models in accurately identifying entities within CSM data. To tackle obstacles, we proposed a NER model CyMapNER for the CSM domain. A clear definition of named entity categories pertinent to the CSM domain was established initially, followed by the creation of a dedicated NER dataset tailored to this domain. Subsequently, we present a domain adaptation training framework that integrates large language models. It combines with data augments, pseudo-labeling and domain-adaptive pretraining to enhance the adaptability of the NER model. The comparative experimental results demonstrate that CyMapNER models outperforms traditional NER models in CSM datasets. The results reveal that by domain adaptation training framework, the recognition accuracy of CyMapNER model reaches 97%, which achieves an improvement from 5.6% to 18.3% over the state-of-the-art NER models, and it performs well in recognizing complex and sparse entities, highlighting its effectiveness in handling the intricacies of CSM data.</p>