基於神經網路之結合資訊壓縮隱藏方法

Abstract

資料隱藏是資訊安全裡重要的研究領域。為了隱藏機密資料的存在，我們將機密資料藏於不相關的圖片中。一般而言，我們希望能藏入大量的資料，並減少對圖片造成的失真。 本篇論文提出一個基於神經網路，並結合資訊隱藏與壓縮的方法。在傳送端，我們用編碼器壓縮圖片，並轉換成二進位資料。接著，我們從二進位資料選出對圖片品質影響較少的部分，並用機密資料取代。在解壓縮階段，接收者可以直接取出隱密質料，並用解碼器來還原圖片。實驗結果顯示出我們能放入更多的機密資料，並保有較佳的還原品質。

Data hiding is an important research field in information security. In order to keep the confidential information from being discovered, we conceal the information in unrelated images. In general, the goal of data hiding is to embed a large amount of information without causing visual distortion. In this thesis, we propose a joint data hiding and compression scheme based on neural networks. On the sender side, we use the encoder to compress the image, and then use the binarizer to generate binary codes. Next, we select parts of the binary codes that have less influence on image quality and replace them with the hidden data. During the decompressing phase, the receiver can extract the hidden data directly and restore the image with the decoder. Experimental results show that the proposed scheme has better hiding capacity and visual quality.