Experimental results.

Steganography, the use of algorithms to embed secret information in a carrier image, is widely used in the field of information transmission, but steganalysis tools built using traditional steganographic algorithms can easily identify them. Steganography without embedding (SWE) can effectively resist detection by steganography analysis tools by mapping noise onto secret information and generating secret images from secret noise. However, most SWE still have problems with the small capacity of steganographic data and the difficulty of extracting the data. Based on the above problems, this paper proposes image steganography without embedding carrier secret information. The objective of this approach is to enhance the capacity of secret information and the accuracy of secret information extraction for the purpose of improving the performance of security network communication. The proposed technique exploits the carrier characteristics to generate the carrier secret tensor, which improves the accuracy of information extraction while ensuring the accuracy of secret information extraction. Furthermore, the Wasserstein distance is employed as a constraint for the discriminator, and weight clipping is introduced to enhance the secret information capacity and extraction accuracy. Experimental results show that the proposed method can improve the data extraction accuracy by 10.03% at the capacity of 2304 bits, which verifies the effectiveness and universality of the method. The research presented here introduces a new intelligent information steganography secure communication model for secure communication in networks, which can improve the information capacity and extraction accuracy of image steganography without embedding.

隐写术（Steganography）指通过算法将秘密信息嵌入载体图像的技术，在信息传输领域应用广泛，但基于传统隐写算法构建的隐写分析工具可轻易识别此类隐写操作。无嵌入隐写术（Steganography Without Embedding, SWE）通过将噪声映射至秘密信息，并由秘密噪声生成秘密图像，可有效抵御隐写分析工具的检测。但多数无嵌入隐写术仍存在隐写数据容量偏小、数据提取难度较大的问题。针对上述问题，本文提出了一种无需嵌入载体秘密信息的图像隐写术，其目标在于提升秘密信息容量与秘密信息提取准确率，以优化安全网络通信的性能。所提方法利用载体特征生成载体秘密张量（carrier secret tensor），在确保秘密信息提取准确性的同时提升信息提取精度。此外，本方法将瓦瑟斯坦距离（Wasserstein Distance）用作判别器的约束项，并引入权重裁剪（Weight Clipping）以进一步提升秘密信息容量与提取准确率。实验结果表明，在2304比特的容量下，所提方法可将数据提取准确率提升10.03%，验证了该方法的有效性与普适性。本研究提出了一种新型智能信息隐写安全通信模型，可应用于网络安全通信场景，能够提升无嵌入图像隐写术的信息容量与提取准确率。