不完全信息环境下指挥控制组织结构识别与发现方法

Nowadays, the Command and Control （C2） organization idetection and detection is one of the most important issues on social networks. The theory analysis about how the organization forms has great value in theory and practice. Especially, the military or terrorist organization which are more antagonistic, can not be observed in time and completely. Therefore they are always under incomplete information and asymmetric environment. The detecting their organization structure rapidly and effecively plays a crucial role in the network-centric warfare. There are a huge number of studies on Command Control information networks, such as gravity analysis of C2 networks, link prediction on C2 relationships and pattern recognition on role and function of the elements in C2 networks. However, it’s hardly to get the complete and unencrypted information so that the many studies can’t be used to predict missing links and analysis on the C2 networks in practice. In this project, the imcomplete information, such as encrypted messages, are studied to reconstruct the C2 networks. The link prediction methods are proposed to identify the missing and false links by the message's basic information itself, such as length of message, sending and receiving addresses, delay and transfer protocols, etc. Based the topology of the C2 network restored by link prediction, we will explore the role recognition for hierarchy network and the structure mining. By this way , the architecture and C2 relationships of enemy can be analyzed for battlefield awareness.

指挥控制组织结构的识别与发现是社会网络研究领域的热点，不仅具有巨大的理论价值而且在实际有着广泛的应用，特别是针对诸如恐怖组织、军事行动等对抗性强的组织，快速准确的指挥控制结构发现是遏制犯罪、保障国家安全的重要环节。然而，这些网络与传统的社会网络不同，他们在对抗过程中往往刻意隐蔽自己的组织结构、传输信息等，使得其具有明显的信息不完备性、部分可观测性及非对称对抗等特点，导致结构识别和发现存在困难。本课题针对强对抗网络，研究不完全信息下的指控网络结构特征，研究结构还原和小组织发现等基础算法，为指控网络实时分析提供理论支撑。

本课题聚焦如何利用指控网络中不完备、非对称数据，实现军事作战组织、恐怖组织等强对抗网络的拓扑还原与关键节点预测，探索指控组织宏观整体效能与指控实体微观变化之间的本质关联与作用规律，开展了链路预测、虚假信息过滤及角色识别问题，主要研究创新点包括四个方面。.（1）针对层级指控网络内部不同类型节点的关联特征，以及指控网络与使命任务的关系，研究层级指控网络的表示，提出了一种基于使命任务的层级指控网络演化模型，便于进行边预测及角色分组；.（2）根据网络中传输的不完整信息，针对层次结构模型提出了一种拓扑还原算法，引入边聚类算法中的分割密度指标，实现链路预测还原程度的度量；.（3）针对大量冗余或虚假信息，基于随机分块理论构建了过滤层级指控网络中虚假边的模型，提出了一种平衡稳定的MH抽样算法实现模型求解；.（4）针对还原后的层级指控网络，依据各类指挥控制节点的业务特点，基于分层最小支配集提出了一种边聚类算法，实现指挥控制节点角色的识别，支撑对层级指控网络的拓扑重构。