- 學位論文
基於WiFi通道狀態資訊與功率延遲特徵之入侵偵測與追蹤技術
Intruder Detection and Tracking Based on Power Delay Profile from WiFi CSI
摘要
基於通道狀態信息（CSI）的入侵偵測系統至今已有許多發展。然而，以往的相關研究在入侵者被阻擋或僅於小範圍內活動時會讓表現變差，並且大多數研究僅偵測入侵者的存在。我們認為一個完善的入侵檢測系統應該足夠穩健、對入侵者敏感，並且能夠在一定程度上追蹤入侵者所在的區域。功率延遲特徵（PDP）根據每條路徑的延遲提供近似路徑分量的資訊。眾所周知，當信號向四面八方傳播時，不同的傳播路徑會不同程度地受到入侵者的影響。基於以上觀點，我們設計了一種演算法能夠尋找最能反映入侵者活動的路徑分量以增加檢測概率，並同時利用PDP提供的延遲資訊來追蹤入侵者位置。首先，我們必須構建一個入侵檢測系統來評估演算法的性能。在以往研究的基礎上，我們設計了一個基礎系統，並在實驗室和教室進行了測試，其平均偵測率能夠達到93.69%。然而，對於一些入侵者被遮擋或遠離檢測設備的位置，偵測率僅為60.77%。經過演算法的增強，遮擋區域的偵測率可以提高到94.44%，整個系統的平均偵測率也提高到98.98%。另外，為了驗證PDP能否達到追蹤的目的，我們還在模擬器上模擬了實驗室環境。我們控制空間中唯一變量為入侵者，在理想環境下使用320MHz頻寬，可以使追蹤誤差小至0.45m。
並列摘要
There have been many developments in intrusion detection systems based on Channel State Information (CSI). However, previous related research performances are limited when intruders are blocked or only move in a small area, and most studies only detect intruders' existence. A sound intrusion detection system should be robust enough, sensitive to intruders, and able to track the area where the intruder is located to a certain extent. Power Delay Profile (PDP) provides approximate path component information based on the delay of each path. It is known that when the signal spreads in all directions, different propagation paths are affected to varying degrees by intruders. Based on the above point of view, we designed an algorithm to find the path component that best reflects the intruder's activities to increase the probability of detection, and at the same time use the delay information provided by the PDP to track the intruder's position. First, we must construct an intrusion detection system to evaluate the algorithm's performance. Based on previous research, we built a basic system and tested it in actual laboratories and classrooms, and its average detection rate has reached 93.69%. However, the detection rate was only 60.77% for some positions where the intruder was obstructed or moved away from the detection devices. After the enhancement of our algorithm, the detection rate of blocked areas could increase to 94.44%, and the average detection rate of the overall system has also risen to 98.98%. Additionally, to verify that PDP can achieve the purpose of tracking, we also simulated the laboratory environment on the simulator. We control the only change in the space is the intruder so that the tracking accuracy can be as small as 0.45m.