近年来，随着数据安全监管的日益收紧，科学数据管理面临越来越严峻的“安全合规”挑战，数据分类分级保护逐渐成为学术界、数据管理实践者和监管机构共同关注的议题。然而，现有的研究和实践大多局限于对数据合规的解释与反应性应对，缺乏对科学数据分类分级保护的系统性和理论性讨论。这种认知不足限制了科学数据安全管理领域理论框架和实用模型的发展。为形成对科学数据分类分级保护的系统性理解，本研究基于对现有实践的广泛调查，提炼出科学数据的六项关键安全特征：多重规制、伦理强规制、学科领域差异性、“规模-风险”帕累托分布、公益性和动态敏感性，以此六项特征为基础，构建了科学数据安全分类和分级框架，并提出了全面、平衡与精简三种保护模式。研究提出了“数据合规-合规成本-数据收益”三角平衡观点，合理解释了三者之间的权衡关系。文中还详细讨论了数据安全分类与安全分级的区别及其相互作用，澄清了科学数据安全分类的复杂性。该研究提出的针对科学数据分类分级保护的理论框架为分析科学数据安全管理中的复杂问题提供了框架性工具，可为相关研究提供有价值的参考，有助于推动科学数据安全保护实践。

In recent years, increasing data security regulations have posed significant compliance challenges for scientific data management. Data classification and grading for protection has become a focal point for academia, practitioners, and regulatory bodies. However, existing research mostly focuses on compliance interpretation and reactive measures, lacking a systematic theoretical analysis of scientific data protection. This gap limits the development of frameworks and models in the field. To address this, based on an extensive survey of current practices, this paper identifies six key security characteristics of scientific data: multi-regulation, strict ethical regulation, disciplinary differences, Pareto distribution of "scale-risk," public interest, and dynamic sensitivity. It proposes a classification and grading framework, along with three protection models: comprehensive, balanced, and streamlined. Additionally, the paper introduces a "compliance-cost-benefit" triangle to explain the trade-offs among these factors. The proposed framework clarifies the complexity of classifying scientific data, distinguishing between data classification and grading, and offering insights into their interaction. This theoretical model provides valuable reference for future research and practical tools for addressing challenges in scientific data security management.