基于机器学习的身份认证协议智能分析与设计方法

本项目面向新兴领域数据/服务资源的规模化及安全访问需求，针对资源访问的类别多样性、交互复杂性、敌手能力动态性等特点，研究基于机器学习和上下文感知的身份认证协议动态形式化分析和智能组合理论与关键技术。首先，通过构建多级综合评价指标体系，提出基于机器学习和上下文感知的身份认证协议情境分类研判模型；其次，结合研判结果，研究不同应用场景中身份认证协议的情境特征与敌手攻击能力间的映射关系，进而构建动态敌手能力简化模型，降低协议分析与设计的复杂性；以此为基础，构建基于着色petri网的身份认证协议动态形式化分析模型。更进一步，以协议组合逻辑的思路为指导，通过对协议形式化描述的静态过程间切片分析，实现协议安全服务组件的分类组织，设计安全组件的动态选择与安全重组策略，提出情境感知的身份认证协议智能组合模型。最终，基于CPN Tools开源工具，搭建情境感知的身份认证协议动态形式化分析与智能化设计原型。

To address the security requirements of service and resource access in the scenario of authentication protocol, the theories and methods of machine learning and context-aware based dynamic formal analysis and intelligent combination of authentication protocol will be studied, for which of the diversity, the interactive complexity, and the ability of the adversary. Firstly, we will find out the essential characteristics of the scenario of authentication protocol. Furthermore, a multistage comprehensive security evaluation system and a machine learning and context-aware based scenario classification model will be proposed. Secondly, along with the result of the model above, heterogeneous characteristics, scale-free features, correlation and business aggregation characteristics of complex network environments will be studied, in order to revel the the mapping between the situational features and the adversary's ability. By this way, a scenario-aware and dynamic formal analysis model can be presented, which has simplified adversary model and reduced calculation. Thirdly, under the guidance of DDMP model and with the static analysis of intraprocedural slicing to the formal description of authentication protocol, the classification of basic security components of the authentication protocol will be constructed. And, the context-aware based intelligent combination of authentication protocol model will be built, in which the dynamic selection and security restructuring strategy of basic security components will be implemented. Finally, an intelligent analysis system of the context-aware based authentication protocol will be designed based on the CPN tools.