面向云端隐私保护的图像隐写方法OA
Image steganography for cloud-based privacy protection
图像隐写在隐蔽通信和云端安全存储中对隐私图像保护具有重要意义.与一次性隐蔽通信不同,云端图像安全存储需要支持隐私图像的反复存取.现有能够实现隐私图像无损重构的隐写方法,虽然能够规避隐私图像在多次隐藏和重构过程中的失真累积问题,但是存在鲁棒性不足的缺陷,难以在有损环境下重构出可感知的隐私图像.为此,提出一种能够在无损环境下实现隐私图像无损重构,同时在有损环境下实现隐私图像可感知重构的方法.首先,提出一种带抖动补偿的嵌套量化索引调制(DC-NQIM)算法,使量化索引调制的嵌入容量翻倍.接着,通过优化载体稳定性、隐私图像位平面与嵌套调制层级之间的协同匹配关系,确保对图像视觉质量贡献更大的位平面被分配至稳定性更好的调制层级.最后,提出一种基于局部结构感知的自适应隐私图像纠错算法,有效避免因位平面误差累积而产生的局部伪点.实验结果表明,所提方法在无损环境下重构隐私图像的峰值信噪比(PSNR)为无穷大,能够支持隐私图像的反复存取;在有损环境下,能够重构出PSNR接近30 dB的可感知隐私图像,以适应有损云端隐私图像保护场景.
Image steganography plays a vital role in protecting private images in covert communication and secure cloud storage.Unlike one-time covert communication,cloud-based secure image storage requires supporting mul-tiple accesses to private images.Existing steganography methods for lossless reconstruction of private images can prevent distortion accumulation over multiple embedding and reconstruction cycles.However,they lack sufficient robustness and struggle to reconstruct perceptually recognizable private images in lossy environments.To address this,a method was proposed that achieves lossless reconstruction of private images in lossless conditions while en-abling perceptually recognizable reconstruction under lossy conditions.First,a nested quantization index modula-tion with dither compensation(DC-NQIM)algorithm was proposed,which doubles the embedding capacity of quantization index modulation.Subsequently,by optimizing the carrier stability and the synergistic matching rela-tionship between the bit-planes of privacy images and the nested modulation levels,it was ensured that the bit-planes contributing more to the visual quality of the image were assigned to the modulation levels with better stabil-ity.Finally,a local structure-aware adaptive privacy images error correction algorithm was introduced to avoid lo-cal artifacts caused by bit-plane error accumulation.Experimental results show that the proposed method achieves an infinite PSNR for reconstructing privacy images under lossless conditions,thereby supporting repeated access to privacy images;under lossy conditions,it can reconstruct perceptually recognizable privacy images with a PSNR close to 30 dB,making it suitable for lossy cloud-based privacy images protection scenarios.
郭波标;杨潇晖;平萍
河海大学计算机与软件学院,江苏 南京 211100河海大学计算机与软件学院,江苏 南京 211100河海大学计算机与软件学院,江苏 南京 211100
信息技术与安全科学
信息隐藏隐私保护以图藏图量化索引调制
information hidingprivacy protectionimage-in-imagequantization index modulation
《网络与信息安全学报》 2026 (4)
17-32,16
中央高校基本科研业务费专项资金资助项目(No.B250205032)国家自然科学基金资助项目(No.61902110)江苏省研究生科研与实践创新计划项目(No.KYCX25_0918)The Fundamental Research Funds for the Central Universities(No.B250205032),The National Natural Science Foundation of China(No.61902110),Postgraduate Research&Practice Innovation Program of Jiangsu Province(No.KYCX25_0918)
评论