Abstract: Aiming at the problem that static access control model is difficult to adapt to data dynamics and context variability in big data environment,this paper proposes a fine-grained access control model based on dynamic data sensitivity level.The model first constructs a multi-dimensional quantitative assessment system to dynamically calculate the real-time sensitivity level of data by analyzing the data content,contextual environment and historical operation behaviors,which overcomes the rigidity of traditional static classification.On this basis,the dynamic sensitivity level is taken as the core decision attribute,and deeply integrated with the attribute-based access control model,a context-adaptive permission dynamic granting and revocation mechanism is designed,which realizes the precise control of different users’ access behaviors at different times,places and scenarios.Experimental results show that the model can effectively perceive the changes in data value and risk while ensuring low performance overhead.Compared with the traditional role-based access control model and the static attribute-based access control model,it significantly improves the accuracy and security of privilege assignment,and it is especially suitable for the big data application scenarios with frequent data flow and changing security requirements,which provides an effective way to build an intelligent and adaptive data security protection system.

Key words: Big data security, Access control, Fine-grained, Dynamic sensitivity levels, Attribute-based access control