项目名称: 基于密集快速特征提取的可视媒体篡改检测研究
项目编号: No.61502241
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 自动化技术、计算机技术
项目作者: 李健
作者单位: 南京信息工程大学
项目金额: 20万元
中文摘要: 本项目将研究图像和视频这些可视媒体的篡改检测方法,可应用于调查取证和网络信息安全等重要领域。研究工作着眼于特征提取这一篡改检测问题的关键点,转变传统思路,不以增加特征维度为目标,而是以加密的方式分析每一个像素点和它周围邻域之间的关系,以更好地发现篡改痕迹,达到提高篡改检测准确率的目的。同时针对视频数据量大的特点,提取特征前不全部解码视频,而是部分解码视频计算得到差分图像,基于此快速提取篡改检测特征。所提取特征的后续处理过程将吸收深度学习领域的研究成果,引入非监督学习、多层稀疏编码等技术,以提高特征分类效果。最后,本项目将在自主搭建的符合实际应用场景的可视媒体库验证理论研究成果的正确性,并开发具有一定实际应用价值的可视媒体篡改检测原型系统。预期研究成果将对网络信息安全和数字取证领域的研究和应用带来贡献。
中文关键词: 篡改检测;统计特征;网络流媒体
英文摘要: This proposal will study the method for detection of tampering in the visible medias, possibly being used for the areas like forensics and Internet information security etc. We focus on the key-point of tampering detection problem, namely feature extraction, to perform our research. Unlike the traditional approach increasing the dimension of the feature vector, we densely extract the features to improve the detection accuracy via utilizing the correlation between the pixel and its neighborhood. Considering the great amount of videos in the Internet, we try to extract features from partly compressed video to guarantee the speed. In addition, inspired by the recent progress in deep learning area, we introduce unsupervised clustering and sparse coding techniques to process the extracted features. The proposal will construct a visible media database which is close to the practical situation, in order to confirm the theoretical result of our research. The expected results will be helpful to the research and application of the areas like Internet information security and digital forensics.
英文关键词: forgery detection;statistical features;Internet stream media