可變性類比網路編碼雙向中繼通道之惡意行為探討

Abstract

在合作式無線網路中，雙向中繼站通道是很基本的構建單元，兩個信源在此通道中透過中繼站交換訊息。先前的文獻探討中，大多假設終端彼此間在中繼站網路中是完全合作的。在本篇論文中，我們考慮的雙向中繼通道包含一個惡意中繼站及一個合作中繼站，而此合作中繼站採用的傳送方式是著名的類比網路編碼。我們的目標是透過相互資訊賽局去了解這些中繼站的最佳策略。在我們所定義的賽局中，參與者是合作式中繼站和惡意中繼站，效用函數定義為可達到的碼率。透過尋找賽局的納許均衡可以得知合作中繼站及惡意中繼站的最佳傳輸策略。因為我們允許中繼站在同一個碼字中的不同符號去調整傳送策略，所以我們提出的賽局比前人的研究更一般化。在這個更新更一般化的賽局中，我們保證了單純策略均衡的存在，而這個單純策略均衡建議合作式中繼站即使是在固定通道下也要去調整不同符號的模擬網路編碼放大係數。然而因為效用函數是非凸函數，要推導出明確的單純策略均衡是非常困難的。為了克服這個困難，我們將效用函數線性化。此外，前人研究中所探討的單向中繼通道是我們論文的一個特殊情況，我們的研究結果顯示前人所提出的惡意攻擊方式不是最佳的。

The two-way relay channel is a fundamental building block of the cooperative wireless networks, where two sources want to exchange messages via relays. Most works in the literature on the relay networks assume that all terminals are fully cooperative. In this thesis, we consider a two-way relay channel that contains a malicious relay, and a cooperative relay which adopts the celebrated analog network coding (ANC). We aim to understand the optimal strategies of these two relays via mutual information games. The cooperative and malicious relays are two players in our game and the achievable sum rates are adopted as the utilities. The optimal transmission strategies of both types of relays are derived by identifying the Nash equilibrium (NE) of the proposed game. By allowing relays to change transmission actions symbol by symbol in a codeword length, the proposed game generalizes the mutual information games in previous works. With this new game formulation, the pure strategy NE exists where the cooperative relay is suggested to vary the amplifying coefficients of the ANC symbol by symbol even for a fixed channel. However, owing to the non-convex nature of our utilities, it is still hard to explicitly derive the pure strategy NE. To overcome this problem, the utilities are linearized. Furthermore, the special cases of our results also show that the malicious attacks proposed by previous works for the one-way relay channel may not be optimal.