基於可攜式文件格式之資訊隱藏技術研究與其應用

Abstract

由於可攜式文件格式（PDF）的各種特性，如開放標準、跨平台及提供高品質的閱覽與列印，讓PDF變得非常普及，也因此PDF變成了一種很好的掩護檔案，可提供秘密通訊之用。在本論文中我們提出一個將秘密資訊編成十進位數字藏入PDF檔案中的方法，可以在很小的誤差範圍下調整文件中的參數，將整數變換為實數，或將實數轉換為另一實數，而讓人無從察覺，是一有相當高欺敵效果的方法。 另一方面，許多公司也將機密文件存成PDF檔案，但在PDF標準中的保密機制並不足以避免惡意使用者散佈這些文件。我們因此提出了一個藏入新型態浮水印及使用者資訊，藉以達到版權保護與檔案散佈控管的方法。我們利用透明圖片作為新浮水印，使其嵌入後不致產生不透明的底色。另外亦將使用者資訊嵌入隨機選出的圖片當中，提高秘密資訊的強靭度，使混淆惡意使用者難於破壞。 PDF檔案非常容易被複製以及修改，因此我們亦提出一透過嵌入驗證訊號來做PDF檔案內容驗證的方法。其驗證訊號由使用者選定的金鑰與以及文字內容所產生，而文件內容即可透過比對驗證訊號，來檢驗文件內容的正確性。 透過好的實驗結果，我們證明了所提出各方法的可行性。

Due to its outstanding capabilities like open-standard, cross-platform, high-quality viewing and printing, the PDF file becomes very popular nowadays. Therefore, the PDF file is a good choice as cover media for covert communication. We propose a method to encode secret messages by converting integers to real numbers with small changes in values, yielding a difference of appearance very difficult to notice by human vision. On the other hand, many companies store their classified documents by the PDF. The methods proposed in the PDF standard are not enough to prevent users from redistributing the classified documents. We propose a method to create a new type of watermark and embed it together with some user information in PDF files for copyright protection. The created watermark has no background color, which comes from mixing a watermark image and an alpha mask. The user information is hidden by a data hide technique in a randomly-selected image so as to confuse illicit users. PDF files are easy to duplicate and modify, so we propose a method for authentication of a PDF file by embedding authentication signals in the text matrices of the file based on the data hiding method we propose for covert communication. The authentication signals are generated from a randomization result of exclusive-ORing the data of the text strings in the PDF file and some random numbers generated with a secret key. The content of the PDF file can be authenticated by matching the extracted authentication signals with the embedded ones. Experimental results show the feasibility of the proposed methods.