可抗幾何攻擊之盲目數位浮水印技術探討

有鑑於數位媒體易於複製與傳播的特性，如何應用浮水印技術保護數位影像的智慧財產權與著作權，一直是重要的研究項目。浮水印的強韌性是評估效能的重要指標，其中又以抵抗幾何攻擊為近年的研究重點。一般而言，幾何攻擊可由旋轉、縮放與位移等3種操作中的一或多種組合而成，目前一般數位浮水印機制，若要具備能抵抗幾何攻擊能力，通常需要配合非盲目或半盲目的浮水印抽取機制，但是當需要保護的數位影像數量持續增加時，此機制將面臨密鑰的管理與查詢問題而難以有效運作。本文整理近期抵抗幾何攻擊的相關研究，並嘗試提出一個結合盲目與抗幾何攻擊浮水印的加密及抽取機制，配合浮水印預設強度以及影像本身統計資訊，結合尺度不變特徵轉換與離散小波轉換，以調適性方式於特定區塊藏入浮水印，實驗結果證實此機制確有其可行性，除可抵抗旋轉及壓縮等攻擊，並能大幅減輕密鑰的存管問題。

關鍵字

盲目浮水印；抗幾何攻擊浮水印；尺度不變特徵轉換；智慧財產權；離散小波轉換

並列摘要

In view of the characteristics of digital media is easy to be replication and propagation, how to apply watermarking to protect the intellectual property and copyright of the digital images has been an important research project. The robustness is an important indicator to evaluate the performance of watermarking systems, among which how to resist geometric attacks is the focus of recent studies. In general, geometric attacks is the combination of operations of rotation, scaling and translation. To resist geometric attacks, most of the digital watermarking systems will require non-blind or semi-blind watermark extraction mechanisms, however these systems will face severe problems of secret keys management and inquiry while the number of protected digital images continuous increasing. In this study we consolidate recent researches of RST invariant watermarking schemes and propose a new system, with the pre-defined watermark strength and statistical information of the protected imagery, and combined with blind watermarking and RST invariant properties. By utilizing the scale-invariant feature transform (SIFT) and discrete wavelet transform, the watermarks can be adaptive embedded into selected blocks. Simulation results show this scheme not only satisfy the requirements to resist common rotation and compression attacks, but also significantly reduce the secret keys depository problems.