Image steganalysis based on CNN and LSTM

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Mar.2023, Vol.46 No.3: 320-325, 391

DOI: 10.3969/j.issn.1003-5060.2023.03.006

Authors

LING Baohong $ ^{1,2} $, ZHENG Gang $ ^{1} $, HU Min $ ^{1} $, PENG Yinyin $ ^{1} $, HU Donghui $ ^{1} $

(1. School of Computer Science and Information Engineering, Hefei University of Technology, Hefei 230601, China; 2. School of Information Engineering, Anhui Broadcasting Movie and Television College, Hefei 230011, China)

Abstract and Keywords

Abstract: To further improve the detection performance of the steganalysis algorithm, this paper proposes a steganalysis algorithm based on convolutional neural network (CNN) and long short-term memory (LSTM). The CNN is used to capture the structural features of cover images, while the LSTM is employed to capture the longer-term temporal feature relations of images in the proposed method. To verify the effectiveness of the hybrid neural network, the proposed method uses XuNet and SRNet as the benchmark steganalysis networks and explores the effective combination of CNN and LSTM. Experimental results show that when combining correctly, the hybrid networks of CNN and LSTM can improve the detection ability compared with the original steganalysis networks.

Keywords: image steganalysis; convolutional neural network (CNN); long short-term memory (LSTM)

作者信息

凌宝红 $ ^{1,2} $，郑钢 $ ^{1} $，胡敏 $ ^{1} $，彭银银 $ ^{1} $，胡东辉 $ ^{1} $

(1. 合肥工业大学计算机与信息学院，安徽合肥 230601；2. 安徽广播影视职业技术学院信息工程学院，安徽合肥 230011)

摘要和关键词

摘要: 为了进一步降低隐写分析算法的检测错误率,文章提出一种基于卷积神经网络(convolutional neural network, CNN)和长短期记忆网络(long short-term memory, LSTM)的隐写分析算法。该算法利用CNN捕获载体图像的结构特征,同时利用LSTM捕获图像的前后时序特征。为了验证混合神经网络的有效性,该算法以XuNet和SRNet为基准隐写分析网络,探讨CNN与LSTM的有效组合方式。实验结果表明,所提方法可以有效提高隐写分析算法的检测能力。

关键词: 图像隐写分析；卷积神经网络（CNN）；长短期记忆网络（LSTM）

Funding

国家自然科学基金资助项目（U1836102）；安徽省科技重大专项资助项目（201903a05020016）；安徽省科研编制计划资助项目（2022AH053074）；高校优秀青年骨干教师国内访问研修资助项目（gxgnfx2020132）和高等学校省级质量工程软件技术专业教学团队资助项目（2021jxtd051）

合肥工业大学学报(自然科学版)

Navigation