Data is the inevitable product of social high informatization, and its security risk is an indispensable part of information security. In the case of accelerating the construction of ubiquitous power Internet of things, this paper deeply studies the data security analysis and label recognition technology based on big data business scenarios. Based on the process of active and passive discovery, a complete catalog of data assets is established by means of database scanning, server scanning, terminal scanning, terminal monitoring, network monitoring and database access monitoring. The threat tree is used to model various threats to data security. On the basis of data asset cataloging and threat modeling, this paper studies the data security risk analysis and quantification method by comprehensively judging the threats faced by the business system, as well as the vulnerability and asset value of the system itself. Based on the heuristic integrated learning strategy of automatic semantic tagging for massive network resources, metadata is extracted through active and passive collection, and the semantic relationship between tags is extracted and processed by natural language processing and machine learning technology. Finally, automatic labels are generated through analysis capabilities such as business correlation analysis, organizational correlation analysis, data distribution analysis, data classification and grade.