作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 安全技术 • 上一篇    下一篇

基于多样性特征的JPEG图像隐写分析

李明则a,向 阳b,张文华c,梁 礼a   

  1. (西安通信学院 a. 研究生管理大队;b. 信息服务教研室;c. 基础部,西安 710106)
  • 收稿日期:2012-12-13 出版日期:2014-01-15 发布日期:2014-01-13
  • 作者简介:李明则(1987-),男,硕士研究生,主研方向:图像隐写分析;向 阳,副教授;张文华、梁 礼,硕士研究生

Steganalysis of JPEG Images Based on Diversity Feature

LI Ming-ze a, XIANG Yang b, ZHANG Wen-hua c, LIANG Li a   

  1. (a. Graduate Management Brigade; b. Department of Information Services; c. Department of Basic, Xi’an Communication Institute, Xi’an 710106, China)
  • Received:2012-12-13 Online:2014-01-15 Published:2014-01-13

摘要: 随着隐写分析技术的发展,新的特征提取算法不断出现,但目前还没有一种较好的通用特征能对JPEG图像进行有效的隐写分析。针对上述问题,提出一种从多域空间提取特征的通用隐写分析算法。采用残差共生矩阵与直方图统计函数计算DCT域、空域、小波域各域系数(像素)之间的依赖性关系,并结合校准方式从中提取特征。对多样性特征维数高的问题,采用前向选择与穷举结合的方法对其降维,以提高分类精度与节约分类时间。对4种典型的JPEG隐写算法在小嵌入率下进行实验,结果表明,与已有的检测方法相比,多域空间提取的多样性特征检测准确率能提高2%以上,适应性更广。

关键词: 隐写分析, 通用特征, 盲检测, 多样性, 小嵌入率, 降维

Abstract: With the development of steganalysis, new features extraction algorithms emerges, but up to now there still has not a set of universal features which can be effective for JPEG images steganalysis. To improve the detection accuracy of blind detection, this paper proposes a universal steganalysis algorithm based on multi-domain features. Markov chains and histogram statistics functions are used to capture the correlations of neighboring coefficients or pixels of DCT domain, spatial domain and wavelet domain as original features. In the same way, the calibrated features are extracted from the calibrated image. Combining boosting feature selections with exhaustion ways, the features improve the detection accuracy and reduce the time of classification after dimensionality reduction. Experiments are done for four kinds of typical JPEG steganography schemes in small embedding rate, compared with existed features, experimental results show that the diversity features get higher accuracy than 2%, and wider adaptability.

Key words: steganalysis, universal feature, blind detection, diversity, small embedding rate, dimensionality reduction

中图分类号: