基於日文字的印表機來源鑑識

Abstract

在相關研究之中，已經有研究使用英文、中文文字做為實驗樣本。本研究使用日文文字做為實驗樣本。因為日文文字非常普遍地出現在日本電視節目、戲劇、電影、動漫遊戲(Anime、Comics and Games，ACG)與智慧型手機應用程式之中，並且這些內容在世界各地的曝光率很高。 隨著資訊科技與網際網路的發展以及數位內容被人們依賴著並且散播在世界各地，數位影像處理相關技術被用在許多領域之中，包括醫學、生物學、工程科學、密碼學、電子商務…等。最普遍的數位影像的輸出裝置不外乎是印表機，然而，利用印表機從事偽造文件、侵害著作權…等議題不斷地出現在新聞媒體中。本研究針對前述的犯罪進行辨識，以數位影像處理技術為基礎，由於日文的字母提供的資訊較少，本研究使用了大量的統計特徵值，分別從灰階共生矩陣、小波離散轉換、高斯濾波器、LoG濾波器、Unsharp濾波器、威能濾波器與賈伯濾波器取得，針對日文文字擷取影像特徵。接著，進一步使用特徵值選擇法來取得較具影響力的特徵值，最後經由支持向量機建立分類模組來進行印表機設備來源之鑑識。

There are already several related researches studying English and Chinese characters for digital forensics. This study uses Japanese characters as the experimental samples instead because Japanese subtitles are appearing in Japanese programs, dramas, movies, Anime, Comic and Games(ACG), smart phone applications frequently over the world. Owing to the advances of new technology, digital contents are widely used over the world. The techniques of digital image processing are also applied to many areas such as Medical Science, Biology, Engineer Science, Cryptography, E-commerce…, etc. Among those applications, the most commonly used device for printing digital content is the printer. Since the criminal counterfeits and copyright infringement is still under study for the news media, there is a need to identify the printer source for Japanese characters. This research offers a new approach to extract the important statistical features from Gray Level Co-occurrence Matrix(GLCM), Discrete Wavelet Transform(DWT), Gaussian filter, Laplacian of Gaussian(LoG) filter, Unsharp filter, Wiener filter and Gabor filter to identify the printer source of unknown printed images for Japanese characters. Subsequently, the feature selection procedures help to get the most important subset of features. Finally, the classification technique of Support Vector Machine(SVM) is utilized to identify the printer source for accurate identification.