信源在中继信道中安全传输的容量问题研究

Characterizing the information theoretical limit of the secrecy capacity for secure source transmission is of theoretical value to design secure transmission schemes with high performance in wireless networks. Secure transmission systems of correlated sources over relay channels cover many issues of wireless communications, such as distributed source-channel coding, cooperative communication and physical-layer security, et cetera. However, exact capacity-equivocation region for secure source transmission is very difficult to be characterized, since uniform source-channel coding and relaying schemes are hard to be constructed in general due to various secrecy requirements of the source information and differences in statistical properties of the side information and channel characteristics. In this project, the secrecy capacity problems for secure transmission of correlated sources over relay channels will be studied from the view of information theory considering different secrecy requirements of the source information. The impacts of the statistical properties of the side information at each node, the correlations of the sources and the channel characteristics on source compression rates and secrecy rates will be studied firstly. Source-channel coding schemes, as well as corresponding relaying strategies will be designed to derive the achievable rate-equivocation regions for the transmission models respectively. Consequently, general inner and outer bounds on the capacity-equivocation regions will be characterized. We will specialize these bounds on the capacity-equivocation regions to the transmission models when the statistical properties of the side information and the channel characteristics meet some certain conditions under which the capacity-equivocation regions are expected to be characterized exactly. With the research on this project, we will try to understand the unique value of distributed source-channel coding and the particular role of cooperation in wireless networks with secrecy constraints, which will be significant in designing secure transmission strategies for secure source transmission over wireless networks.

从信息论角度刻画信源安全传输的容量极限，是无线网络信息安全传输效能得以充分发挥的关键。信源在中继信道中安全传输的通信模型，完美地表征了无线网络中分布式信源-信道编码、协作通信和物理层安全等特征。然而，因信源信息对各类节点的保密性需求不同、节点边信息统计特性和信道传输特性的差异，难以构造统一的信源-信道编码方案和中继协作策略，致使刻画确切的容量-疑义度区域非常困难。为此，本项目以相关信源在中继信道中的安全传输为背景，针对三种典型的传输模型，从信息论角度研究节点边信息的统计特性、信源间的相关性、信道传输的随机性等因素对信源压缩码率和安全速率的影响，构建信源-信道编码方案和中继协作策略，建立容量-疑义度区域外界的求解方法，刻画容量-疑义度区域的一般性内界和外界，并试图得到节点边信息和信道特性满足特定条件时容量-疑义度区域的确切表示。本项研究可为无线网络中设计安全的传输方案提供理论依据和参考。

从信息论角度刻画信道安全传输的容量极限从理论上证明了无线信道安全传输性能的极限，针对无线信道特征和不同通信场景优化收发器的设计是实际通信系统能够达到或接近信息安全传输极限重要措施之一。为此，本项目首先研究了双向脏纸中继信道的容量区域，针对信道中节点具有不同的边信息的三种不同情况，基于晶格码和计算-转发中继策略，分刻画了容量区域的内界，并证明了所得到的容量区域内界与割集外界间的间隙为一常数。其次，研究了无线携能安全通信最优收发器的设计，针对外部窃听节点被动窃听信息和主动攻击两种情况，建立了能量波束成形向量、信息波束向量等参数在内的联合优化模型，通过半定松弛等方法将非凸的优化问题转换为凸优化问题进行求解，并设计了相应的优化算法，给出了收发器最优设计方案，通过数值仿真验证了优化方案的有效性。同时，将机器学习方法应用到无线安全通信物理层优化设计中。针对中继-窃听信道中如何选择最优中继节点以保证最大的信息安全传输速率，将中继选择问题建模为多分类问题，提出了一种基于深度神经网络的中继选择方案，该方案能够得到传统最优中继选择方案接近的性能，但与传统中继选择方案相比，其仅需一半的通信反馈开销。最后，针对多用户干扰信道发送节点的功率分配问题，提出了一种基于递归神经网络的功率分配方法，该方案不仅以相当低的复杂度实现了与传统的基于加权最小均方误差算法的功率分配方案几乎相近的加权和速率，并且以较少模型参数开销获得了比基于深度神经网络的功率分配方案更高的系统和速率。本项目的研究成果可以为无线网络设计最优的传输方案提供一定的理论参考和应用借鉴。