面向网络系统的一致性安全隐私分析与防护机制设计

Consensus is a fundamental distributed estimation, control and optimization algorithm in network systems, and its applications have covered any very critical network systems, e.g., Wireless Sensor Networks, Smart Grid, Social Networks, etc. However, due to the openness and intelligence of network systems, the potential security and privacy issues of the consensus algorithm have been gradually attracting the public awareness. It has become a significant research subject how to guarantee both security and privacy for the consensus algorithm in a distributed way such that it can be further developed. This project is aiming to design the security and privacy preserving algorithms, and establish a complete theory of the consensus security and privacy. Then, we focus on the following three aspects: 1) Studying the machine learning based recognition technology against the multimodal attack; 2) Designing the security and privacy preserving distributed network system architecture based on Topology Hiding, Data Encryption, and Coordination Monitoring; 3) Breaking the theoretical bottleneck, stablishing a complete theory of security and privacy of consistency and building a novel evaluation system based on information theory and stochastic theory. The project is expected to make breakthroughs in the consensus security and privacy and multimodal attack recognition technology, which provides an effective theoretical basis and critical technology support for the implementation of network systems security and privacy.

一致性是网络系统中分布式估计、控制和优化的基础算法，其应用已覆盖网络系统中诸多关键领域，如无线传感器网络、智能电网、社会网络等。随着网络系统开放性和智能性的不断增强，一致性算法在加强安全和隐私方面的需求日益突出，如何实现该算法的分布式安全防御和隐私保护，成为其应用推广亟待克服的挑战性问题之一。本项目以设计确保一致性和安全隐私性的算法为核心，以构建一致性安全和隐私防御机制为目标，开展以下三方面研究：1）基于机器学习理论，研究多模态攻击行为建模和分布式辨识技术；2）基于拓扑隐藏、信息加密和协同互监等技术，设计确保一致性的分布式安全防御和隐私保护机制；3）基于信息熵理论和统计学理论设计有效的安全隐私性评估和分析方法，突破分布式安全隐私性评估的理论瓶颈。本项目有望在一致性安全隐私理论和多模态攻击行为认知技术方面实现突破，为实现网络系统的安全高效运行提供重要理论分析基础和关键技术支持。

一致性是网络系统中分布式估计、控制和优化的基础算法，其应用已覆盖网络系统中诸多关键领域，如无线传感器网络、智能电网、社会网络等。随着网络系统开放性和智能性的不断增强，一致性算法在加强安全和隐私方面的需求日益突出，如何实现该算法的分布式安全防御和隐私保护，成为其应用推广亟待克服的挑战性问题之一。本项目以设计确保一致性和安全隐私性的算法为核心，以构建一致性安全和隐私理论框架为目标。通过一年的努力，得到以下成果：1）给出了随机加噪隐私保护机制下保证一致性收敛的充分性和必要性条件，为算法设计提供理论支撑；2）创新性的提出了(ϵ,δ)数据隐私量化概念，克服差分隐私的不足，并给出该量化指标下的最优估计以及参数之间的解析关系；3）设计并证明了均匀噪声添加机制是最优的一致性隐私防护机制。 本项目在一致性安全隐私理论方面实现了突破，为实现网络系统的安全高效运行提供重要理论分析基础和关键技术支持。