视频隐藏分析关键技术研究

发布时间：2018-11-19 13:39

【摘要】：作为大容量信息隐秘通信的有效手段,视频信息隐藏技术引起人们的关注。为了有效检测以视频为载体的有害隐秘通信行为,视频隐藏分析具有重要的研究价值和应用前景。本文研究以数字视频为载体的隐藏分析技术,针对多种作用域的隐藏算法提出了相应的检测算法,.并将领域适应与通用隐藏检测相结合,分析其在解决载体分布失配问题上的适用性。本文主要工作与创新包括以下6个方面：1、在压缩视频空域隐藏分析方面,提出将时域相关性和空域相关性相结合的SPEAM统计特征,用于检测压缩视频的空域扩频隐藏。在非压缩视频的隐藏检测性能方面,该特征与只考虑空域相关性的图像隐藏分析特征相当,且优于现有视频隐藏分析特征；但在压缩视频的检测性能方面,该特征优于已有的视频隐藏分析特征和图像隐藏分析特征,且具有较低的运算复杂度。2、在压缩域运动矢量隐藏分析方面,针对现有运动矢量隐藏检测方法或者检测性能差且无法适应多种视频编码格式,或者因重压缩参数与原始编码参数不同而失效的缺点,提出基于运动矢量局部最优假设的AoSO特征。该特征在检测性能上明显优于现有的两类特征,且适用范围较广,在多种复杂视频编码条件下均表现出较好的检测性能。3、在H.264帧内预测模式隐藏分析方面,结合H264视频压缩编码标准,提出以SATD作为编码代价的帧内预测模式校准机制及其相应的CIPM隐藏检测特征。该特征在隐藏检测性能上明显优于现有特征,且计算复杂度比较低。4、在改善载体分布失配条件下的隐藏分析算法性能方面,针对其中的隐藏算法失配问题,对领域适应中的特征迁移算法——TCA算法开展研究,并将TCA算法用于解决隐藏算法失配条件下的运动矢量隐藏检测问题。最后通过实验验证其可能带来失配条件下隐藏检测性能的提升,并结合实验中遇到的问题,对TCA等领域适应算法的适用性进行了探讨,并指出领域适应算法在隐藏检测应用中的局限性。5、在视频空域隐藏软件MSU StegoVideo的隐藏分析方面,.针对现有隐藏分析方法仅考虑隐藏信息的检测且检测性能不理想的问题,对MSU StegoVideo隐藏的三个指纹特征进行了分析,包括特殊的颜色空间转换公式、棋盘状掩膜图案以及嵌入强度控制机制,提出基于MSU StegoVideo指纹的非压缩视频隐藏检测与隐藏参数估计方法——MP算法。该算法在检测性能上远优于已有算法,且可对隐藏参数"noise level"以及隐藏信息序列进行初步的估计。6、在格式隐藏软件OpenPuff的隐藏分析方面,针对该软件支持的MPEG、VOB、 FLV、MP4、3GP等格式视频,结合相应的视频容器标准,通过分析隐藏操作引入的文件差异,提出相应的隐藏分析方法,能够实现高精度的盲检测与部分隐藏信息的提取。
[Abstract]:As an effective means of large capacity information secret communication, video information hiding technology has attracted people's attention. In order to detect the harmful hidden communication behavior with video as the carrier, video hiding analysis has important research value and application prospect. In this paper, the hiding analysis technology based on digital video is studied, and the corresponding detection algorithm is proposed for various scope hiding algorithms. Combining domain adaptation with general hidden detection, its applicability in solving mismatch of carrier distribution is analyzed. The main work and innovations of this paper are as follows: 1. In the analysis of compressed video spatial domain hiding, a SPEAM statistical feature combining temporal correlation with spatial domain correlation is proposed to detect spatial spread spectrum hiding of compressed video. In the aspect of hiding detection performance of uncompressed video, this feature is comparable to that of image hiding analysis which only considers spatial correlation, and is superior to the existing video hiding analysis feature. However, in the detection performance of compressed video, this feature is superior to the existing features of video hiding analysis and image hiding analysis, and has lower computational complexity. 2. Aiming at the shortcomings of the existing motion vector hiding detection methods, such as poor detection performance and being unable to adapt to various video coding formats, or because the recompression parameters are different from the original coding parameters, The AoSO feature based on the local optimal assumption of motion vector is proposed. This feature is obviously superior to the existing two kinds of features in detection performance, and has a wide range of applications. It has good detection performance under various complex video coding conditions. 3. In H.264 intra prediction mode hiding analysis, Based on the H264 video compression coding standard, an intra prediction mode calibration mechanism with SATD coding cost and its corresponding CIPM hidden detection features are proposed. This feature is obviously superior to the existing features in the performance of hiding detection, and its computational complexity is relatively low. 4. In order to improve the performance of the hidden analysis algorithm under the mismatch condition of carrier distribution, the proposed algorithm mismatches the hidden algorithm. The feature migration algorithm (TCA) in domain adaptation is studied, and the TCA algorithm is used to solve the problem of motion vector hiding under the condition of mismatch. Finally, it is verified by experiments that it may bring about the improvement of the performance of hidden detection under mismatch conditions, and the applicability of adaptive algorithms in TCA and other fields is discussed in combination with the problems encountered in the experiments. It also points out the limitation of domain adaptation algorithm in the application of hiding detection. 5. In the aspect of hiding analysis of video spatial hiding software MSU StegoVideo. Aiming at the problem that the existing hidden analysis methods only consider the detection of hidden information and the detection performance is not satisfactory, three fingerprint features hidden by MSU StegoVideo are analyzed, including the special color space conversion formula. The chessboard mask pattern and embedding intensity control mechanism are discussed. An uncompressed video hiding detection and parameter estimation method based on MSU StegoVideo fingerprint is proposed, which is called MP algorithm. The algorithm is much better than the existing algorithm in detection performance, and can estimate the hidden parameter "noise level" and the hidden information sequence. 6. In the aspect of hiding analysis of format hiding software OpenPuff, the MPEG,VOB, FLV, supported by the software can be analyzed. Based on the video container standard of MP4,3GP and the corresponding video container standard, the paper puts forward the corresponding hidden analysis method by analyzing the file difference caused by hiding operation, which can achieve high precision blind detection and partial hidden information extraction.
【学位授予单位】：解放军信息工程大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN919.81

【参考文献】