網際網路自1970年代發展至今，儼然已成為現代人生活中不可或缺的一部分，各式各樣不同的網路協定同時提供著不同的應用服務，例如：發布和接收HTML網頁服務由過往的HTTP（Hypertext Transfer Protocol）到因資安意識抬頭而延伸出的HTTPS（HTTP Scure）、提供電子郵件服務的SMTP、POP3及IMAP、提供檔案傳輸的FTP等，而具備網路傳輸路由決定性的路由傳輸協定，由於資安意識抬頭，近年開始受到關注。 本文探討基於機器學習的方式，計算網路路由協定所建構的「路由路徑（Route）」、或「路由圖（Graph）」間的「相似度（Similarity computation）」，並藉由與這些Graph相同路由來源的傳輸內容（例如：特定傳輸特徵、簽章指標），探詢路由協定與傳輸內容間的關聯，即完成「從相同路由來源找尋未知但值得關注的簽章指標」及「已知重要簽章指標存在其他路由的可能性」等兩個重要任務。實驗過程中我們另外利用「Graph Theory」將路由協定傳輸路徑與節點視覺化，使我們在取得相似度計算結果後得輔以視覺化的路由圖，快速瞭解行經節點及簽章指標間的關聯。 經由實驗，我們證明透過基於神經網路的SimGNN模型可以協助我們取得路由圖之間的相似度，包含Graph Edit Distance（GED）及相似度分析計算，進而達成上述兩個重要的實驗目的，對於未來資安防護及簽章指標查找溯源具有應用及研究前景。 ;Since its development in the 1970s, the Internet has become an indispensable part of modern life. Various network protocols provide different application services, such as HTTP （Hypertext Transfer Protocol）for publishing and receiving HTML web pages, HTTPS（HTTP Secure）for enhanced security, SMTP, POP3, and IMAP for email services, and FTP for file transfer. Routing protocols, such as Border Gateway Protocol（BGP） play a crucial role in determining network transmission routes. However, their security implications have gained attention in recent years. This paper explores the application of machine learning techniques to compute the similarity between routes or graphs constructed by routing protocol. By analyzing the transmission content with the same route source, such as specific transmission features or indicators, we investigate the relationship between routing protocols and transmission content. This approach aims to identify unknown but noteworthy feature indicators based on similar route sources and explore the possibility of known important feature indicators existing in other routes. Additionally, we visualize the routing protocol′s transmission paths and nodes to quickly understand the associations between traversed nodes and feature indicators after obtaining similarity computation results. Through experimentation, we demonstrate that the SimGNN（Similarity Graph Neural Network）model based on neural networks can assist in obtaining similarity between routing graphs, including Graph Edit Distance（GED）and similarity analysis calculations. This achievement fulfills the two important experimental objectives mentioned earlier and holds potential for future applications and research prospects in the fields of cybersecurity defense and feature indicator traceability.
