使用多天線之物理層安全技術

Abstract

本論文主要探討多天線之物理層的安全傳輸，並包含兩個部分。在第一部分中，我們提出了一個多輸入多輸出預先編碼與後置編碼技術結合的收發機架構，此架構可達到物理層的資料安全傳輸。當發送機可以知道授權接收機的完全通道資訊時，所提出的預先編碼可以使得接收機端的訊號雜音比最大化，並達到安全傳輸。當接收機只能獲得授權接收機的部分通道資訊時，我們修正了洛伊德演算法，進而架構一個編碼書去量化預先編碼。在這一部分，我們也將說明，我們所提出的編碼書不會影響到所提出之架構的傳輸安全性。除此之外，當完全的通道資訊不能獲得時，我們提出了一個後置編碼的方法，降低訊號雜音比的衰減。我們也進一步地從還原率與傳輸安全性的觀點下，分析如何所提出的架構之效能。從分析結果顯示，提出的架構可以如何達到完全還原率並達到完美安全傳輸。最後，數值模擬的結果與理論分析的結果呈現一致，並且展現當不同的還原演算法 使用的情況下，所提出架構的不同優勢。

在第二部分，我們研究，對於多輸入單一輸出單一竊聽天線的通道並且發送機只能獲得授權接收機的部分通道資訊下，如何量化遮罩波束成形的系統去達到最大化的安全傳輸率。在此種情況下，由於通道資訊的量化，導致人造雜訊漏回至授權接收機。在此方面的研究，使用全部的量化位元去量化波束成形向量。則被量化的波束成形向量之零空間被使用去傳送人造雜訊。我們發現此類的量化架構會導致授權接收機端的嚴重干擾。為了克服這個議題，我們提出量化波束成形向量與人造雜訊向量應該被分開化。因此提出的量化方式，會有兩個編碼書，從第一個編碼中，選取一個波束成形向量去最大化波束成形增益：從第二個編碼書中，選取一個人造雜訊向量去減低干擾。從理論分析的結果呈現，分開量化的方式可以顯著地降低授權接收機端的干擾。進一步，基於提出量化的架構，我們去分析如何分配量化位元去達到最大安全傳輸率。藉由所提出的量化方式與為元 分配方式，安全傳輸的效能相對於傳統量化的架構可以顯著改善。數值模擬結果也與理論分析的結論一致。

This dissertation consists of two parts. In the ?rst part, we propose a MIMO precoding and postcoding transceiver to achieve data secrecy at the physical layer. When full channel state information (CSI) of the legitimate receiver is known to the transmitter, the proposed precoder can simultaneously maximize receive SNR and attain secrecy. When only partial CSI is available, a modi?ed Lloyd algorithm is proposed to construct codebooks for quantizing the precoder. As will be explained in the article, the use of the proposed codebooks does not affect the secrecy of the proposed sys- tem. Moreover, we propose a low-complexity postcoder to compensate the SNR loss when full CSI is not available. Furthermore, the performance of the proposed system is analyzed from the view points of both recovery rate and secrecy. Based on the analyzed results, we show how to achieve full recovery rate and perfect secrecy for the proposed system. Simulation results corroborate the theoretical results, and show that the proposed system enjoys different advantages when different recovery algorithms are used.

In the second part, we investigates how to quantize the masked beamforming systems to maximize the secrecy rate for MISOSE (multiple-input, single-output, multiple-eavesdropper) channels, when only partial channel state information (CSI) at the legitimate receiver is available to the transmitter. In this case, the artificial noise (AN) leaks to the legitimate receiver due to CSI quantization. In the literature, all quantization bits are used to quantize the beamforming vector. Then the null space of this quantized beamforming vector is used to transmit the AN. We ?nd that such quantization schemes can result in serious interference at the legitimate receiver. To overcome this issue, we propose that the beamforming vector and the AN vector should be quantized separately, where the beamforming vector should be selected from a codebook to maximize the beamforming gain and the AN vector should be selected from another codebook to minimize the leakage (or interference). Theoretical results show that separate quantization can significantly reduce the AN leakage at the legitimate receiver. Furthermore, based on the proposed quantization scheme, we show how to allocate bits to separately quantize the beamforming vector and the AN vector to maximize the secrecy rate. By using the proposed quantization and bit allocation schemes, the secrecy rates of masked beamforming systems can be improved compared to the conventional quantization schemes. Simulation results corroborate the theoretical results.