Text Steganalysis Method Based on Hierarchy-Aware Matching

doi:10.19678/j.issn.1000-3428.0069385

Abstract

Abstract: Existing text steganalysis models experience difficulty in learning and extracting multilayer effective information that truly exists in encrypted data. To address this issue, a text steganalysis method, HAM-Stega, based on hierarchy-aware matching is proposed. This method utilizes the matching relationship between the relative distance between text information and label information in steganographic data to obtain a feature-matching relationship between text and coarse- and fine-grained labels in a hierarchy-aware manner. Based on this, joint embedding and matching learning loss functions are designed to guide the classification of text feature representations and obtain the final hierarchical classification information. The experimental results show that HAM-Stega's detection accuracy on the Large multidistribution mixed dataset, which is similar to real-world scenarios, improves by approximately 1.25—7.42 percentage points compared to the comparison model, indicating that the proposed model has an effective steganalysis detection capability on mixed datasets. Simultaneously, HAM-Stega can extract and detect other layers of effective information present in the steganographic data, such as steganographic algorithms for encrypted text, embedding rates, and corpus types. It improves the hierarchical classification metrics Macro-F1 and Micro-F1 by 5.41 and 4.36 percentage points, respectively, compared with the pretrained BERT model.

Key words: information security, text steganalysis, hierarchy-aware matching, Graph Neural Network (GNN), BERT

摘要： 针对现有文本隐写分析模型难以学习和提取载密数据中真实存在的多层有效信息的问题,提出一种基于层次感知匹配的文本隐写分析方法HAM-Stega。该方法利用隐写数据中的文本信息与标签信息之间相对距离的匹配关系,以层次感知的方式获取文本与粗粒度、细粒度标签之间的特征匹配关系,以此设计联合嵌入损失函数和匹配学习损失函数,引导文本特征表示进行分类学习,得到最终的层次分类信息。实验结果表明,HAM-Stega在更符合现实场景的多分布混合数据集Large上的检测精度比对比模型提高了1.25~7.42百分点,表明该模型在混合数据集上具有有效的隐写分析检测能力。同时, HAM-Stega对于隐写数据中存在的其他多层有效信息(载密文本的隐写算法、嵌入率、语料类型等)可以进行提取和检测,其在层次分类指标Macro-F1和Micro-F1上相较于预训练的BERT模型分别提高了5.41和4.36百分点。

关键词: 信息安全, 文本隐写分析, 层次感知匹配, 图神经网络, BERT

CLC Number:

TP391

JIA Jianghao, ZHANG Ziwei, GAO Liting, WEN Juan, XUE Yiming. Text Steganalysis Method Based on Hierarchy-Aware Matching[J]. Computer Engineering, 2026, 52(2): 245-252.

贾江浩, 张梓葳, 郜丽婷, 文娟, 薛一鸣. 基于层次感知匹配的文本隐写分析方法[J]. 计算机工程, 2026, 52(2): 245-252.

/ Recommend / Download Citations

URL: https://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0069385

https://www.ecice06.com/EN/Y2026/V52/I2/245

References

[1] GUO L, NI J, SHI Y Q. Uniform embedding for efficient JPEG steganography[J]. IEEE Transactions on Information Forensics and Security, 2014, 9(5): 814-825.
[2] HOLUB V, FRIDRICH J. Designing steganographic distortion using directional filters[C]//Proceedings of the IEEE International Workshop on Information Forensics and Security (WIFS). Washington D.C.,USA:IEEE Press,2013: 234-239.
[3] 卞玉星, 黄荣, 周树波, 等. 基于可逆神经网络的多载体图像隐写模型[J]. 计算机工程, 2024, 50(12): 213-223. BIAN Y X, HUANG R, ZHOU S B, et al. Multi-cover image steganography model based on invertible neural network[J]. Computer Engineering, 2024, 50(12): 213-223. (in Chinese)
[4] HUANG Y F, TANG S Y, YUAN J. Steganography in inactive frames of VoIP streams encoded by source codec[J]. IEEE Transactions on Information Forensics and Security, 2011, 6(2): 296-306.
[5] YANG Z L, PENG X S, HUANG Y F. A sudoku matrix-based method of pitch period steganography in low-rate speech coding[EB/OL].[2024-01-05]. https://link.springer.com/chapter/10.1007/978-3-319-78813-5_40.
[6] 张晓虹, 项世军, 黄红斌. 利用可逆网络的音频藏图算法[J]. 西安电子科技大学学报, 2024, 51(4): 226-238. ZHANG X H, XIANG S J, HUANG H B. Hiding images in audio based on invertible neural networks[J]. Journal of Xidian University, 2024, 51(4): 226-238. (in Chinese)
[7] LUO Y B, HUANG Y F. Text steganography with high embedding rate: using recurrent neural networks to generate Chinese classic poetry[C]//Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security. New York,USA:ACM Press,2017: 99-104.
[8] YANG B Y, PENG W L, XUE Y M, et al. A generation-based text steganography by maintaining consistency of probability distribution[J]. KSII Transactions on Internet & Information Systems, 2021, 15: 4184-4202.
[9] 向凌云, 杨双辉, 王蓉, 等. 基于序列到隐写序列的约束型自然语言信息隐藏方法[J]. 计算机学报, 2023, 46(8): 1650-1669. XIANG L Y, YANG S H, WANG R, et al. Constrained linguistic steganography based on sequence to steganographic sequence model[J]. Chinese Journal of Computers, 2023, 46(8): 1650-1669. (in Chinese)
[10] 杨雨, 张梓葳, 文娟. 基于胶囊网络的多任务少样本文本隐写分析[J]. 计算机学报, 2022, 45(12): 2592-2604. YANG Y, ZHANG Z W, WEN J. Multi-task few-shot text steganalysis based on capsule network[J]. Chinese Journal of Computers, 2022, 45(12): 2592-2604. (in Chinese)
[11] XIANG L Y, SUN X M, LUO G, et al. Research on steganalysis for text steganography based on font format[C]//Proceedings of the 3rd International Symposium on Information Assurance and Security. Washington D.C.,USA:IEEE Press,2007: 490-495.
[12] CHEN Z L, HUANG L S, MIAO H B, et al. Steganalysis against substitution-based linguistic steganography based on context clusters[J]. Computers & Electrical Engineering, 2011, 37(6): 1071-1081.
[13] MENG P, HANG L S, YANG W, et al. Linguistic steganography detection algorithm using statistical language model[C]//Proceedings of the International Conference on Information Technology and Computer Science. Washington D.C.,USA:IEEE Press,2009: 540-543.
[14] WEN J, ZHOU X J, ZHONG P, et al. Convolutional neural network based text steganalysis[J]. IEEE Signal Processing Letters, 2019, 26(3): 460-464.
[15] NIU Y, WEN J, ZHONG P, et al. A hybrid R-BiLSTM-C neural network based text steganalysis[J]. IEEE Signal Processing Letters, 2019, 26(12): 1907-1911.
[16] YANG H, BAO Y J, YANG Z L, et al. Linguistic steganalysis via densely connected LSTM with feature pyramid[C]//Proceedings of the 2020 ACM Workshop on Information Hiding and Multimedia Security. New York,USA:ACM Press,2020: 5-10.
[17] YANG J S, YANG Z L, ZHANG S Y, et al. SeSy: linguistic steganalysis framework integrating semantic and syntactic features[J]. IEEE Signal Processing Letters, 2022, 29: 31-35.
[18] WU H Z, YI B, DING F, et al. Linguistic steganalysis with graph neural networks[J]. IEEE Signal Processing Letters, 2021, 28: 558-562.
[19] GUO S N, LIU J Y, YANG Z L, et al. Linguistic steganalysis merging semantic and statistical features[J]. IEEE Signal Processing Letters, 2022, 29: 2128-2132.
[20] XU Q, ZHANG R, LIU J Y. Linguistic steganalysis by enhancing and integrating local and global features[J]. IEEE Signal Processing Letters, 2023, 30: 16-20.
[21] GUO S N, LIU J Y, YANG Z L, et al. Linguistic steganalysis merging semantic and statistical features[J]. IEEE Signal Processing Letters, 2022, 29: 2128-2132.
[22] TASKIRAN C M, TOPKARA U, TOPKARA M, et al. Attacks on lexical natural language steganography systems[J]. Security, Steganography, and Watermarking of Multimedia Contents VIII, 2006, 6072: 607209.
[23] XIANG L Y, SUN X M, LUO G, et al. Linguistic steganalysis using the features derived from synonym frequency[J]. Multimedia Tools and Applications, 2014, 71(3): 1893-1911.
[24] XUE Y M, KONG L Z, PENG W L, et al. An effective linguistic steganalysis framework based on hierarchical mutual learning[J]. Information Sciences, 2022, 586: 140-154.
[25] GO A, BHAYANI R, HUANG L. Twitter sentiment classification using distant supervision[EB/OL].[2024-01-05]. https://cs.stanford.edu/people/alecmgo/papers/TwitterDistant Supervision09.pdf.
[26] MAAS A L, DALY R E, PHAM P T, et al. Learning word vectors for sentiment analysis[EB/OL].[2024-01-05]. https://ai.stanford.edu/~ang/papers/acl11-WordVectors SentimentAnalysis.pdf.
[27] FANG T N, JAGGI M, ARGYRAKI K. Generating steganographic text with LSTMs[EB/OL].[2024-01-05]. https://arxiv.org/abs/1705.10742.
[28] YANG Z L, GUO X Q, CHEN Z M, et al. RNN-stega: linguistic steganography based on recurrent neural networks[J]. IEEE Transactions on Information Forensics and Security, 2019, 14(5): 1280-1295.
[29] YANG Z L, ZHANG S Y, HU Y T, et al. VAE-stega: linguistic steganography based on variational auto-encoder[J]. IEEE Transactions on Information Forensics and Security, 2021, 16(4): 880-895.
[30] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[EB/OL].[2024-01-05]. https://arxiv.org/abs/1810.04805.

Please choose a citation manager

Content to export