利用可逆网络的音频藏图算法

doi:10.19665/j.issn1001-2400.20240303

摘要/Abstract

摘要：

可逆网络因其具有天然可逆的结构,非常适用于信息隐藏领域。图像能以生动直观、有层次的方式传递信息,而音频是一种广泛传播和使用的媒体文件,具有较大的嵌入容量,因此在音频中隐藏图像具有较高的研究和应用价值。在音频藏图任务中,如何表征音频和图像数据以及如何在减少音频失真的同时提高重建图像的质量是两个重要的问题。针对这两个问题,提出了一种基于可逆网络的音频藏图算法。对于数据特征表示,受到JPEG图像压缩中数据处理方法的启发,提出了图像特征提取与表示模块,该模块对彩色图像依次进行分块离散余弦变换、锯齿扫描和高低频分离操作,提取出图像的频域特征并得到其一维表示。此外,为了减少音频失真并提高重建图像的质量,利用小波变换分离音频的高低频分量并引入可逆网络将秘密图像嵌入到载体音频的高频区域中。实验结果表明,所提出的算法在实现高嵌入率的同时,能生成质量更高的隐写音频以及重建出更加还原的彩色图像,且算法具有较高的安全性。

关键词: 隐写, 图像隐藏, 可逆网络, 小波变换, 离散余弦变换

Abstract:

Invertible Neural Networks(INNs) are well suited for the field of information hiding due to the fact that their inherent reversible structure.Images are able to efficiently convey information in a vivid and hierarchical manner,while audio is a widely used and distributed media file with a large embedding capacity.Therefore,hiding images in audio is of high research and application value.In the task of hiding images in audio,how to represent audio and image data and how to improve the quality of reconstructed images while reducing audio distortion are two important issues.To address these two problems,this paper proposes an algorithm based on INNs to hide images in audio.Inspired by the data processing methods in JPEG image compression,an image feature extraction and representation module is proposed for data feature representation.This module performs block-wise discrete cosine transform,Zigzag scanning,and high-low frequency separation operation on color images,extracting the frequency domain features of the image and obtaining its one-dimensional representation.In addition,in order to reduce audio distortion and improve the quality of reconstructed images,this paper uses the wavelet transform to separate the high and low frequency components of audio and introduces INNs to embed the secret image into the high-frequency region of the cover audio.Experimental results show that the proposed algorithm can generate higher quality steganographic audio and reconstruct more restored color images while achieving a high embedding rate,and that the proposed algorithm exhibits good security.

Key words: steganography, image hiding, invertible neural networks, wavelet transforms, discrete cosine transforms

中图分类号:

TP309.7

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 Xiaohong, XIANG Shijun, HUANG Hongbin. Hiding images in audio based on invertible neural networks[J]. Journal of Xidian University, 2024, 51(4): 226-238.

图/表 11

图1

图2

图3

图4

表1

表2

表3

图5

图6

图7

表4

参考文献 36

[1]	韩烨, 关晴骁, 刘年生, 等. 一种基于局部曲率的空域图像隐写失真代价[J]. 西安电子科技大学学报, 2023, 50(4):229-236.
	HAN Ye, GUAN Qingxiao, LIU Niansheng, et al. Distortion Scheme Based on the Local Curvature for Spatial Image Steganography[J]. Journal of Xidian University, 2023, 50(4):229-236.
[2]	范文同, 李震宇, 张涛, 等. 基于隐写噪声深度提取的JPEG图像隐写分析[J]. 西安电子科技大学学报, 2023, 50(4):157-169.
	FAN Wentong, LI Zhenyu, ZHANG Tao, et al. JPEG Image Steganalysis Based on Deep Extraction of Stego Noise[J]. Journal of Xidian University, 2023, 50(4):157-169.
[3]	GERA A, VYAS V. Hiding Capacity and Audio Steganography Model Based on LSB in Temporal Domain[J]. Recent Patents on Engineering, 2022, 16(2):65-74.
[4]	NEMATOLLAHI M A, AL-HADDAD S A R, ZARAFSHAN F. Blind Digital Speech Watermarking Based on Eigen-Value Quantization in DWT[J]. Journal of King Saud University-Computer and Information Sciences, 2015, 27(1):58-67.
[5]	GEETHAVANI B, PRASAD E V, ROOPA R. A New Approach for Secure Data Transfer in Audio Signals Using DWT[C]//2013 15th International Conference on Advanced Computing Technologies(ICACT).Piscataway:IEEE, 2013: 1-6.
[6]	BAZIYAD M, SHAHIN I, RABIE T, et al. 64-Bit Quantization:Taking Payload Capacity of Speech Steganography to the Limits[J]. Multimedia Tools and Applications, 2022, 81(28):40561-40579.
[7]	VARGHESE F, SASIKALA P. Secure Data Transmission Using Optimized Cryptography and Steganography Using Syndrome-Trellis Coding[J]. Wireless Personal Communications, 2023, 130(1):551-578.
[8]	KREUK F, ADI Y, RAJ B, et al. Hide and Speak:Deep Neural Networks for Speech Steganography(2019)[J/OL].[2019-02-07].https://arxiv.org/abs/1902.03083v1.
[9]	JIANG S, YE D, HUANG J, et al. SmartSteganogaphy:Light-Weight Generative Audio Steganography Model for Smart Embedding Application[J]. Journal of Network and Computer Applications, 2020,165:102689.
[10]	岳峰, 朱慧, 苏兆品, 等. 基于BN优化SNGAN的自适应音频隐写[J]. 计算机学报, 2022, 45(2):427-440.
	YUE Feng, ZHU Hui, SU Zhaopin, et al. An Adaptive Audio Steganography Using BN Optimizing SNGAN[J]. Journal of Computers, 2022, 45(2):427-440.
[11]	ZHANG R, DONG H, YANG Z, et al. A CNN Based Visual Audio Steganography Model[C]//International Conference on Adaptive and Intelligent Systems. Heidelberg:Springer, 2022:431-442.
[12]	GELETA M, PUNTI C, MCGUINNESS K, et al. Pixinwav:Residual Steganography for Hiding Pixels in Audio[C]//ICASSP 2022-2022 IEEE International Conference on Acoustics,Speech and Signal Processing(ICASSP).Piscataway:IEEE, 2022: 2485-2489.
[13]	ASAD M, GILANI J, KHALID A. An Enhanced Least Significant Bit Modification Technique for Audio Steganography[C]//International Conference on Computer Networks and Information Technology. Piscataway:IEEE, 2011:143-147.
[14]	MISHRA A, JOHRI P, MISHRA A. Audio Steganography Using ASCII Code and GA[C]//2017 International Conference on Infocom Technologies and Unmanned Systems(Trends and Future Directions)(ICTUS).Piscataway:IEEE, 2017:646-651.
[15]	GANWANI P, GUPTA L, JAIN C, et al. Lsb Based Audio Steganography Using Rsa and Chacha20 Encryption[C]//2021 12th International Conference on Computing Communication and Networking Technologies(ICCCNT).Piscataway:IEEE, 2021: 1-6.
[16]	MALIK H M A, ANSARI R, KHOKHARA A. Robust Data Hiding in Audio Using Allpass Filters[J]. IEEE Transactions on Audio,Speech,and Language Processing, 2007, 15(4):1296-1304.
[17]	ERFANI Y, SIAHPOUSH S. Robust Audio Watermarking Using Improved TS Echo Hiding[J]. Digital Signal Processing, 2009, 19(5):809-814.
[18]	WANG S, YUAN W, UNOKI M. Multi-Subspace Echo Hiding Based on Time-Frequency Similarities of Audio Signals[J]. IEEE/ACM Transactions on Anizhenudio,Speech,and Language Processing, 2020,28:2349-2363.
[19]	FALLAHPOUR M, MEGíAS D. High Capacity Method for Real-Time Audio Data Hiding Using the FFT Transform[C]//Advances in Information Security and Its Application:Third International Conference,ISA 2009.Heidelberg:Springer, 2009:91-97.
[20]	TAKAHASHI N, SINGH M K, MITSUFUJI Y. Source Mixing and Separation Robust Audio Steganography(2021)[J/OL].[2021-10-11]. https://arxiv.org/abs/2110.05054v1.
[21]	WU J, CHEN B, LUO W, et al. Audio Steganography Based on Iterative Adversarial Attacks Against Convolutional Neural Networks[J]. IEEE Transactions on Information Forensics and Security, 2020,15:2282-2294.
[22]	CUI W, LIU S, JIANG F, et al. Multi-Stage Residual Hiding for Image-Into-Audio Steganography[C]//ICASSP 2020-2020 IEEE International Conference on Acoustics,Speech and Signal Processing(ICASSP).Piscataway:IEEE, 2020: 2832-2836.
[23]	PAUL S, MISHRA D. Hiding Images within Audio Using Deep Generative Model[J]. Multimedia Tools and Applications, 2023, 82(4):5049-5072.
[24]	XIAO M, ZHENG S, LIU C, et al. Invertible Image Rescaling[C]//Computer Vision-ECCV 2020:16th European Conference.Heidelberg:Springer, 2020:126-144.
[25]	RONNEBERGER O, FISCHER P, BROX T. U-Net:Convolutional Networks for Biomedical Image Segmentation[C]//Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015:18th International Conference.Heidelberg:Springer, 2015:234-241.
[26]	刘婷, 任延珍, 王丽娜. 基于条件可逆网络的生成式图像隐写算法[J]. 信息安全学报, 2023, 8(4):17-30.
	LIU Ting, REN Yanzhen, WANG Lina. Generative Image Steganography via Conditional Invertible Neural Network. Journal of Cyber Security, 2023, 8(4):17-30.
[27]	JING J, DENG X, XU M, et al. HiNet:Deep Image Hiding by Invertible Network[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Piscataway:IEEE, 2021:4733-4742.
[28]	LU S P, WANG R, ZHONG T, et al. Large-Capacity Image Steganography Based on Invertible Neural Networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE, 2021:10816-10825.
[29]	LIU Y, QIN Z, ANWAR S, et al. Invertible Denoising Network:A Light Solution for Real Noise Removal[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE, 2021:13365-13374.
[30]	FANG H, QIU Y, CHEN K, et al. Flow-Based Robust Watermarking with Invertible Noise Layer for Black-Box Distortions[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto: AAAI, 2023:5054-5061.
[31]	LUO Y, ZHOU T, LIU F, et al. IRWArt:Levering Watermarking Performance for Protecting High-Quality Artwork Images[C]//Proceedings of the ACM Web Conference 2023. New York: ACM, 2023:2340-2348.
[32]	FONSECA E, PLAKAL M, ELLIS D P W, et al. Learning Sound Event Classifiers from Web Audio with Noisy Labels[C]//ICASSP 2019-2019 IEEE International Conference on Acoustics,Speech and Signal Processing(ICASSP).Piscataway:IEEE, 2019: 21-25.
[33]	LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft Coco:Common Objects in Context[C]//Computer Vision-ECCV 2014: 13th European Conference.Heidelberg:Springer, 2014:740-755.
[34]	KINGMA D P, BA J. Adam:A Method for Stochastic Optimization(2014)[J/OL].[2014-12-22]. https://arxiv.org/abs/1412.6980v6.
[35]	YANG J, ZHENG H, KANG X, et al. Approaching Optimal Embedding in Audio Steganography with GAN[C]//ICASSP 2020-2020 IEEE International Conference on Acoustics,Speech and Signal Processing(ICASSP).Piscataway:IEEE, 2020: 2827-2831.
[36]	杨晓元, 唐洪琼, 钮可, 等. 一种基于宏块复杂度的视频隐写算法[J]. 西安电子科技大学学报, 2022, 49(2):164-172.
	YANG Xiaoyuan, TANG Hongqiong, NIU Ke, et al. Video Steganography Based on Macroblock Complexity[J]. Journal of Xidian University, 2022, 49(2):164-172.

网络结构	操作	输出尺寸
输入特征		c_in×32 768
卷积单元1	Conv1d+PReLU	32×16 384
卷积单元2	Conv1d+PReLU	64×8 192
卷积单元3	Conv1d+PReLU	128×4 096
卷积单元4	Conv1d+PReLU	256×2 048
转置卷积单元1	deConv1d+PReLU拼接卷积单元3的输出	128×4 096
转置卷积单元2	deConv1d+PReLU拼接卷积单元2的输出	64×8 192
转置卷积单元3	deConv1d+PReLU拼接卷积单元1的输出	32×16 384
转置卷积单元4	deConv1d	c_out×32 768

方法	嵌入容量/(Kb·s^-1)
ISGAN^[11]	529.2
PixInWav^[12]	1 027.3
文中方法	1 058.4

方法	隐写音频			重构图像
方法	SNR/dB	MSE	MOS	PSNR/dB	SSIM	MOS
ISGAN^[11]	-0.08	3.73E-4	1.58	17.69	0.6913	3.29
PixInWav^[12]	17.34	1.32E-5	3.42	27.73	0.9019	4.17
文中sw/o dct	31.30	4.86E-7	4.36	11.48	0.3204	1.00
文中sw/o dwt	19.72	1.27E-5	4.06	23.14	0.6351	2.94
文中方法	20.30	1.23E-5	4.13	30.99	0.9065	4.48

方法	检测准确率/%
ISGAN^[11]	97.56
PixInWav^[12]	82.13
文中方法	59.88