【作者】 杨博；

【导师】 郭立；

【作者基本信息】 中国科学技术大学 ， 电路与系统， 2010， 硕士

【摘要】 隐写和隐写分析是信息隐藏技术两个主要领域,隐写(Steganography)是将秘密信息隐藏到载体中,隐写分析(Steganalysis)是对隐写的攻击,目的是检测隐秘通信中的秘密信息,破坏秘密通信。近年来,随着信息隐藏技术的发展,出现了以音乐设备数字接口(MIDI)音频为载体的隐写和隐写分析方法。MIDI是音乐设备和计算机之间传输音乐信息的协议标准,MIDI文件存储的是如何播放音乐的指令,具有容量小,便于编辑等优点,在手机及互联网领域有广泛应用。所以针对MIDI音频的信息隐藏技术研究具有重要意义。由于MIDI音频存储的是MIDI指令,MIDI信息隐藏技术与图像载体及其他音频载体隐藏既有相似之处又有不同特点。所以本文在国家自然科学基金的资助下,针对MIDI载体的信息隐藏技术进行研究,论文的主要工作和创新点如下:1分析音频隐写的机理,总结音频隐写和隐写分析方法,提出了较为完善音频隐写和隐写分析评价模型,总结目前MIDI音频的隐写方法和隐写分析方法,按照评价模型比较各方法的优劣。2目前只有针对MIDI音频LSB替换的隐写分析方法,本文针对MIDI音频LSB匹配隐写,提出基于力度平滑度转换率的隐写分析算法,利用匹配隐写改变MIDI力度平滑度,训练平滑度转换率阈值,判断MIDI音频是否存在隐写;并根据匹配隐写的加性噪声特性,给出嵌入率的一种估计方法。实验表明,在嵌入率大于20%时,检测正确率可以达到70%以上,对嵌入率的估计也比较准确,估计偏差在10%之内。3为了提高MIDI音频LSB隐写分析检测效率,提出基于特征提取和模式识别的隐写分析算法。提取直方图特征函数(HCF)域21维统计矩特征组成特征向量,用支持向量机(SVM)训练分类器。实验表明,分别对LSB替换、匹配隐写进行分析,平均分类正确率可以达到90%以上。并利用MIDI音频分通道存储的特性,给出了特征提取的优化方法,经过实验分析,优化分析算法的分类正确率和适用性均有提高。4针对MIDI音频LSB、省略指令和同步指令三种隐写方法,分析隐写后MIDI文件结构特征和统计特征,提出一种MIDI隐写分析的盲分类器的设计方法。实验表明,盲分类器可以实现MIDI音频三种隐写方法的有效检测。根据MIDI音频LSB隐写和隐写分析方法,实现了一个简单在线MIDI音频隐写和分析系统,系统采用浏览器服务器(B/S)模式,能够实现远程在线对MIDI音频的LSB隐写和隐写分析。更多还原

【Abstract】 Steganography and steganalysis are two main areas of Information hiding technology. Steganography is to hide secret information into the carrier. Steganalysis purposed is to detect and destroy the secret information is an attack on steganography. In recent years, with the development of information hiding technology, MIDI audio is being the carrier hidden secret message. MIDI file with the advantage of small size and editing easy is stored on the instructions of how to play music. MIDI is a music information transmission protocol standard between Music equipment and computers, which is widely used in the field of mobile phones and the internet. As a result, MIDI audio information hiding technique is more important.MIDI audio is stored as MIDI commands. There are both similarities and differences between MIDI file information hiding and other carieers. This paper mainly research on MIDI audio information hiding supported by National Natural Science Foundation, thesis work and innovations are as follows:1 Summary audio steganography and steganalysis method, proposed a more complete audio steganography and steganalysis evaluation model. Summarize the MIDI audio steganographic methods and steganalysis methods according to the evaluation model.2 As there is only MIDI audio LSB replacement Steganalysis method, a method for MIDI audio LSB matching is proposed in the paper. The method judge a MIDI audio file by training smoothness conversion rate threshold, as the process of Steganography would change velocity smoothness of the MIDI file. The method can return a probable embedding rate based on additive noise feature of the steganography. Experiments show that when the embedding rate greater than 20%, the correct detection rate can reach more than 70%, also the error of the embedding rate is less than 10%.3 In order to improve MIDI audio LSB detection efficiency, a steganalysis algorithm is proposed based on the characteristic mining and pattern recognition. Extract 21-dimensional statistical characteristics of histogram characteristic function (HCF) field, training classifier with the support vector machine (SVM). Experiments show that the average classification accuracy rate can be above 90% for either LSB replacement or matching steganography. Finally, propose a LSB steganalysis optimization feature extraction methods using the characteristic of MIDI audio sub-channel storage. Experiments show that both classification accuracy and applicability increased.4 A MIDI blind steganalysis classifier was designed by Analyzing MIDI files structural features and statistical features for MIDI audio LSB steganography, omitted instructions steganography and synchronization instructions steganography. Experiments show that the blind classifier can detect three methods steganography of MIDI audio effectively. Finally, design a simple online steganography and steganalysis system using the browser and server (B/S) mode, which can achieve LSB steganography and Steganalysis of MIDI audio online.更多还原