应用科学学报 ›› 2016, Vol. 34 ›› Issue (6): 670-676.doi: 10.3969/j.issn.0255-8297.2016.06.003

• 信号与信息处理 • 上一篇    下一篇

基于LSB检测的JPEG隐写分析特征增强方法

郑国华, 冯国瑞, 余江, 程航, 张新鹏   

  1. 上海大学通信与信息工程学院, 上海 200444
  • 收稿日期:2015-08-17 修回日期:2015-11-17 出版日期:2016-11-30 发布日期:2016-11-30
  • 通信作者: 张新鹏,教授,博导,研究方向:信息隐藏、加密信号处理,E-mail:xzhang@shu.edu.cn E-mail:xzhang@shu.edu.cn
  • 基金资助:

    国家自然科学基金(No.61525203,No.61472235,No.61373151,No.U1536109);上海市曙光计划基金(No.14SG36);上海市优秀学术带头人计划基金(No.16XD1401200)资助

JPEG Steganalysis Based on LSB Detection and Enhanced Features

ZHENG Guo-hua, FENG Guo-rui, YU Jiang, CHENG Hang, ZHANG Xin-peng   

  1. School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China
  • Received:2015-08-17 Revised:2015-11-17 Online:2016-11-30 Published:2016-11-30

摘要:

隐写分析构造共生矩阵特征时因截断而导致部分信息丢失,于是提出基于DCT系数最低有效位面的特征扩展方法,增加对截断处嵌入信息的描述. 对DCT系数平面取模生成最低有效位平面,通过统计相邻位置的取值,生成二阶、三阶共生矩阵. 为进一步增强特征的性能,引入Zigzag扫描,获得重新排列的DCT系数平面,在此基础上提取相应特征. 通过上述方法,在原有LIU特征的基础上构造508维特征. 结果表明,在对抗DCT系数最低有效位嵌入的隐写中,所提方法能有效提高对隐写图像的检测率.

关键词: JPEG隐写分析, LSB平面, Zigzag扫描

Abstract:

A method of feature enhancement based on detection of least significant bits of DCT coefficients is proposed to solve the problem of information loss due to truncation operation in the co-occurrence matrix. The DCT coefficient plane is calculated modulotwo to generate the least significant bit plane. Co-occurrence matrices are used to capture correlated features among neighboring coefficients of the least significant bit plane. To enhance the features for detection, Zigzag scan is used to produce a new DCT coefficient plane, and then a co-occurrence matrix is generated based on the new plane. A feature vector with 508 dimensions is extracted using the LIU method. Experimental results show that performance of the extended features is better than those of previous schemes.

Key words: LSB plane, JPEG steganalysis, Zigzag scan

中图分类号: