西安电子科技大学学报 ›› 2023, Vol. 50 ›› Issue (4): 229-236.doi: 10.19665/j.issn1001-2400.2023.04.022

• 网络空间安全专栏 • 上一篇    下一篇

一种基于局部曲率的空域图像隐写失真代价

韩烨1(),关晴骁1(),刘年生1(),陈和风1(),张卫明2(),高岩3()   

  1. 1.集美大学 计算机工程学院,福建 厦门 361021
    2.中国科学技术大学 网络空间安全学院,安徽 合肥 230026
    3.中国电子学会,北京 100000
  • 收稿日期:2022-11-07 出版日期:2023-08-20 发布日期:2023-10-17
  • 通讯作者: 关晴骁
  • 作者简介:韩烨(1998—),女,集美大学硕士研究生,E-mail:786928913@qq.com;|刘年生(1967—),男,教授,E-mail:nsliu@jmu.edu.cn;|陈和风(1982—),女,副教授,E-mail:chenhf@jmu.edu.cn;|张卫明(1976—),男,教授,E-mail:zhangwm@ustc.edu.cn;|高岩(1983—),男,高级工程师,E-mail:gaoyanem@139.com
  • 基金资助:
    国家自然科学基金(U1936114);福建省自然科学基金项目(2021J01857);福建省软科学项目(2020R0066);海南省重大科技计划(ZDKJ2019003);中国国家铁路集团有限公司科技研究开发计划项目(N2021W003)

Distortion scheme based on the local curvature for spatial image steganography

HAN Ye1(),GUAN Qingxiao1(),LIU Niansheng1(),CHEN Hefeng1(),ZHANG Weiming2(),GAO Yan3()   

  1. 1. School of Computer Engineering,Jimei University,Xiamen 361021,China
    2. School of Cyberspace Security,University of Science and Technology of China,Hefei 230026,China
    3. Chinese Institute of Electronics,Beijing 100000,China
  • Received:2022-11-07 Online:2023-08-20 Published:2023-10-17
  • Contact: Qingxiao GUAN

摘要:

隐写术是一种隐蔽通信方法。随着人们越来越青睐于社交媒体的信息交流,以图片为载体的信息隐藏技术是当前研究的热点。在最小化嵌入失真的框架下,为了提高隐写术的安全性,定义良好的代价函数至关重要。针对该问题,提出了一种基于二阶差分构建Hessian矩阵计算局部二阶方向梯度残差计算空域自适应代价的方法。首先利用横向与纵向的二阶差分,以及四个方向的差分构建Hessian矩阵。对于各个方向的Hessian矩阵,计算特征值的绝对值作为残差,并采用多种方法在多个阶段融合残差与多方向代价。实验以SRM和MaxSRMd2作为特征,以及集成FLD分类器作为隐写分析方法来评估性能。实验对比了8种方法,且每种方法重复10次。实验数据表明,在相同的嵌入率下,使用提出的失真代价方案会表现出更高的样本检测错误率。提出的方法与当前主流的空域隐写代价Hill方法性能相近,在与Hill代价采用争议像素优先融合方法后,可进一步提升空域自适应隐写代价的安全性。

关键词: 隐写术, 隐写分析, 加性失真, 代价函数, 失真函数

Abstract:

Steganography is an information hiding method.As people increasingly favor social media for information exchange,information hiding technology with images is a hot topic of current research.In the framework of minimizing embedding distortion,it is crucial to define a good cost function in order to improve the security of steganography.In this paper we propose a distortion scheme for spatial steganography,in which the Hessian matrix is constructed via the second order difference and is used for calculating the residual.First,we use the horizontal difference and vertical difference,and directional differences in 4 directions,to calculate Hessian matrices in 4 directions.For these Hessian matrices,we calculate absolute values of their eigenvalues as a residual.Along with cost values calculated from residuals,we merge them in two phases.In the experiment,we use two kinds of steganalysis feature sets,SRM and MaxSRMd2,respectively,to evaluate the performance of each distortion scheme.Moreover,an ensemble FLD classifier is adopted to map the steganalysis feature to the prediction results for each testing sample.The experimental result is compared to those of eight methods and each method is repeated ten times.Experimental data show that using the distortion cost scheme proposed in this paper exhibits a higher sample detection error rate at the same embedding rate.Experimental results show that the proposed method achieves a security performance comparable to that by the Hill.In addition,when our proposed method is fused with the Hill by the Controversial Pixel Prior (CPP) rule,its performance can be enhanced.

Key words: steganography, steganalysis, additive distortion, cost function, distortion function

中图分类号: 

  • TP309.7