| 引用本文： |
-
关晴骁,朱杰,赵险峰,于海波,刘长军.一种基于线性规划特征选择和集成分类器的图像隐写分析方法[J].信息安全学报,2018,3(1):83-94 [点击复制]
- GUAN Qingxiao,ZHU Jie,ZHAO Xianfeng,YU Haibo,LIU Changjun.Image Steganalysis Based on Linear Programing Feature Selection and Ensemble Classifier[J].Journal of Cyber Security,2018,3(1):83-94 [点击复制]
|
|
| 本文已被：浏览 6210次 下载 5720次 |
码上扫一扫！ |
| 一种基于线性规划特征选择和集成分类器的图像隐写分析方法 |
|
关晴骁1,2, 朱杰1,2, 赵险峰1,2, 于海波1,2, 刘长军1,2
|
|
|
| (1.中国科学院信息工程研究所信息安全国家重点实验室, 北京 中国 100093;2.中国科学院大学网络空间安全学院, 北京 中国 100093) |
|
| 摘要: |
| 隐写分析是防范由隐写术进行信息隐藏所带来危害的有效方法。图像隐写分析方法主要用于检测图像是否被隐写术嵌入隐秘信息。通用型图像隐写分析能够针对广泛类型的隐写术进行检测，该类方法一般采用从图像提取的统计特征和分类器模型进行。当前的高性能隐写分析一般采用高维特征和集成分类器进行。高维特征能够较好地表达图像统计特性中被隐写术扰动的成分，但另一方面，高维特征具有较多的冗余和无效成分，因此进行特征选择能较好的提升效率。本文提出一种使用线性规划的特征选择模型，该模型可与集成分类器协同使用，同时考虑集成分类器中子分类器的检测精度和多个子分类器使用特征的多样性。实验证明，本文提出的方法对多个隐写术的检测性能有较好的提升。 |
| 关键词: 隐写分析 特征选择 隐写术 信息隐藏 |
| DOI：10.19363/j.cnki.cn10-1380/tn.2018.01.006 |
| 投稿时间：2016-06-14修订日期：2016-10-20 |
| 基金项目:本课题得到国家自然科学基金（U1536105，U1636102），国家重点研发计划（2016QY15Z2500，2016QY15Z2500）资助。 |
|
| Image Steganalysis Based on Linear Programing Feature Selection and Ensemble Classifier |
|
GUAN Qingxiao1,2, ZHU Jie1,2, ZHAO Xianfeng1,2, YU Haibo1,2, LIU Changjun1,2
|
| (1.State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China;2.School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100093, China) |
| Abstract: |
| steganalysis is an effective method to prevent the vicious usage of steganography. Image steganalysis can detect the presence of secret message embedded by steganography in digital image. Universal image steganalysis method is designed to detect various kinds of steganography, such methods usually use ensemble classifier and high dimensional feature which can capture the disturbance introduced by steganography embedding. On another hand, there are many ineffective and redundant components in high dimensional feature, thus feature selection methods can enhance its detection accuracy. In this paper we propose a feature selection method for high dimensional feature and ensemble classifier based image stegnalysis. In this method, we consider the accuracy of base classifiers in ensemble classifier and the diversity of subsets of feature used by them. Experimental result shows that our method can improve detection performance on many kinds of steganography. |
| Key words: Steganalysis Feature selection Steganography Information hiding |
