【正文】
學校代號 10532 學 號 S1010W139 分 類 號 TP391 密 級 普 通 工程碩士學位論文 空域圖像LSB匹配隱寫分析技術(shù) 學位申請人姓名 田清龍 培 養(yǎng) 單 位 信息科學與工程學院 導師姓名及職稱 吳蓉暉 副教授 學 科 專 業(yè) 計算機應用技術(shù) 研 究 方 向 網(wǎng)絡(luò)與信息安全 論文提交日期 2012年 3月8日 學校代號:10532學 號:S1010W139密 級:普通湖南大學工程碩士學位論文空域圖像LSB匹配隱寫分析技術(shù)學位申請人姓名: 田清龍 導師姓名及職稱: 吳蓉暉 副教授 培 養(yǎng) 單 位: 信息科學與工程學院 專 業(yè) 名 稱: 計算機應用技術(shù) 論文提交日期: 論文答辯日期: 答辯委員會主席: The Research on Steganalysis of LSB Matching in Spatial Domain of ImagesbyTIAN Qinglong. (Hunan University) 2009A thesis submitted in partial satisfaction of the Requirements for the degree ofMaster of ScienceinComputer Science and Technology in the Graduate schoolof Hunan UniversitySupervisorAssociate Professor Wu RonghuiMarth, 2012工程碩士學位論文湖 南 大 學學位論文原創(chuàng)性聲明本人鄭重聲明:所呈交的論文是本人在導師的指導下獨立進行研究所取得的研究成果。除了文中特別加以標注引用的內(nèi)容外,本論文不包含任何其他個人或集體已經(jīng)發(fā)表或撰寫的成果作品。對本文的研究做出重要貢獻的個人和集體,均已在文中以明確方式標明。本人完全意識到本聲明的法律后果由本人承擔。作者簽名: 日期: 年 月 日學位論文版權(quán)使用授權(quán)書本學位論文作者完全了解學校有關(guān)保留、使用學位論文的規(guī)定,同意學校保留并向國家有關(guān)部門或機構(gòu)送交論文的復印件和電子版,允許論文被查閱和借閱。本人授權(quán)湖南大學可以將本學位論文的全部或部分內(nèi)容編入有關(guān)數(shù)據(jù)庫進行檢索,可以采用影印、縮印或掃描等復制手段保存和匯編本學位論文。本學位論文屬于保密163。,在______年解密后適用本授權(quán)書。不保密R。(請在以上相應方框內(nèi)打“P”)作者簽名: 日期: 年 月 日導師簽名: 日期: 年 月 日摘 要隨著互聯(lián)網(wǎng)技術(shù)的不斷普及和多媒體信息的數(shù)字化,打破了傳統(tǒng)的時空觀念,人們可以迅速的將信息以各種形式傳播到世界的任何角落,但也暴露了越來越重要的安全問題。隱寫術(shù)利用數(shù)字媒體的感知冗余和數(shù)據(jù)冗余,將秘密信息以一定的編碼方式隱藏于數(shù)字媒體中,彌補了密碼學技術(shù)的局限性,被廣泛應用于涉及國家安全、商業(yè)機密、個人隱私等各種信息的安全傳輸。然而隱寫術(shù)也往往被犯罪集團、邪教組織等傳遞非法信息和指令,嚴重威脅國家安全和社會穩(wěn)定。隱寫分析技術(shù)是對隱寫術(shù)的攻擊,阻止隱寫術(shù)被非法利用,對維護國家安全和社會穩(wěn)定有著重要的意義。數(shù)字圖像是因特網(wǎng)中最為常見的數(shù)字媒體,冗余度大,非常適合隱蔽信息,以數(shù)字圖像為載體的隱寫技術(shù)成果最豐富、最成熟,應用也最廣泛。因此對數(shù)字圖像隱寫分析技術(shù)進行深入研究非常必要。本文主要研究空域圖像隱寫分析技術(shù),主要研究成果如下:(1) 通過對LSB匹配隱寫算法進行建模,提出了一種基于圖像直方圖幾何度量的數(shù)字圖像隱寫分析特征。首先將LSB匹配模擬為在圖像中加入隨機噪聲,在圖像直方圖上,LSB匹配相當于對圖像直方圖進行低通濾波,導致圖像直方圖被平滑。曲率是刻畫平滑最有效的方式,因此通過直方圖曲率和來刻畫直方圖變平滑的現(xiàn)象。為了克服圖像內(nèi)容多樣性對隱寫分析造成的影響,采用二次嵌入進行特征校準。在大規(guī)模圖像庫上,使用支持向量機(SVM)進行訓練和測試,實驗結(jié)果表明,該特征具有很高的檢測率,性能優(yōu)于其他同類算法。(2) 提出了一類基于曲率模式矩陣和馬爾科夫鏈相結(jié)合的高維圖像隱寫分析特征。首先分析隱寫算法對圖像像素相關(guān)性的影響,然后采用非線性曲率計算公式對圖像數(shù)據(jù)進行非線性變換,在非線性變換域?qū)ζ溥M行馬爾科夫鏈的建模,得出超高維的隱寫分析特征。為了避免特征維數(shù)過高帶來的維數(shù)災難等問題,采用基于ROC的特征選擇技術(shù),得到適合于隱寫分析的低維特征子空間。使用Ensemble分類器在大規(guī)模圖像庫上進行訓練和測試,實驗結(jié)果顯示,該算法實驗性能優(yōu)于目前主流的隱寫分析算法。關(guān)鍵詞:LSB匹配;隱寫術(shù);隱寫分析;數(shù)字圖像;機器學習AbstractWith the gradual popularization of the Internet technology and the digitization of the multimedia information, people can spread information to any corner of the world quickly through various forms, which breaks the tradition of spatiotemporal concepts. But it also exposes an increasingly important security issues. Steganography hides the secret information into the digital media in some encoded mode by using the perceputual redundancy and data redundancy of digital media, which makes up for the limitation of the cryptography technology, and it is widely used in transmitting some information such as national security, trade secrets, privacy and so on. However, steganography is often used to transmit illegal information and instruction by criminal groups, cults and so on, which threatens national security and social stability seriously. Steganalysis attempts to break steganography, and blocks steganography to be used illegally. Steganalysis is of great significance to safeguard the national security and social stability.As the most mon digital media in Internet, digital image with big redundancy is very fit for hiding information. The research result of steganography with digital image as the carrier is the most abundant and skilled, which is the most widely used. So the further research for digital image steganalysis is very necessary. This paper mainly studies steganalysis on spatialdomain image, the main research results are as follows:(1) By modeling LSB matching based image steganography techniques, we proposed a dectection method based on geometric measures of image histogram. First, LSB matching can be modeled as adding independent additive noise to the image, this will lead to image histogram smoothed by a low pass filter. Curvature is the best way to measure smoothness, and is utilized to evaluate the smoothness of the histogram. Then, the calibration mechanism based secondary steganogtaphy is introduced to reduce the steganalytic difficulty caused by the image variety. SVM are utilized to train and test the classifiers on large image databases, Experimental results show that the proposed method is efficient to detect the LSB matching steganography and has superior results pared with the same kind of other algorithms.(2) A highdimensional feature space for steganalysis of LSB matching is proposed based on curvature mode matrix and markovchain. First, we analysis the impact of LSB matching to dependences between pixels in nature images, then we get the curvature mode matrix by nonlinear curvature transformation to the data of image and model the curvature mode matrix using a markov chain to get the the highdimensional feature space. A feature selection algorithm based on receiver operating characteristic (ROC) analysis is introduced to obtain the feature subspace which suit to steganalysis. Ensemble Classifiers are utilized to train and test the classifiers on large image databases and experimental results show that the proposed method outperforms stateoftheart techniques.Key Words:LSB matching。 stegnanlysis。 steganography。 digital image。 machine learning目 錄湖南大學學位論文原創(chuàng)性聲明 I摘 要 IIAbstract III目 錄 V插圖索引 VII附表索引 VIII第1章 緒 論 1 研究背景與意義 1 研究現(xiàn)狀 2 筮待解決的問題 4 本文主要工作 5 本文結(jié)構(gòu) 5第2章 圖像隱寫及隱寫分析技術(shù) 7 數(shù)字圖像隱寫技術(shù) 7 數(shù)字圖像隱寫分析技術(shù) 10 隱寫分析數(shù)學模型 10 隱寫分析性能指標 11 經(jīng)典的隱寫分析技術(shù) 12 小結(jié) 15第3章 基于圖像直方圖幾何度量的LSB匹配檢測 17 引言 17 特征提取 18 LS