SemSteDiff: Generative Diffusion Model-based Coverless Semantic Steganography Communication

Abstract

Semantic communication (SemCom), as a novel paradigm for future communication systems, has recently attracted much attention due to its superiority in communication efficiency. However, similar to traditional communication, it also suffers from eavesdropping threats. Intelligent eavesdroppers could launch advanced semantic analysis techniques to infer secret semantic information. Therefore, some researchers have designed Semantic Steganography Communication (SemSteCom) schemes to confuse semantic eavesdroppers. However, the state-of-the-art SemSteCom schemes for image transmission rely on the pre-selected cover image, which limits the generalization. To address this issue, we propose a Generative Diffusion Model-based Coverless Semantic Steganography Communication (SemSteDiff) scheme to hide secret images into generated stego images. The semantic related private and public keys enable legitimate receiver to decode secret images correctly while the eavesdropper without the completely correct key-pairs fail to obtain them. Simulation results demonstrate the effectiveness of the plug-and-play design in different Joint Source-Channel Coding (JSCC) frameworks. Results under different eavesdropping settings show that, when Signal-to-Noise Ratio (SNR) = 0 dB, the peak signal-to-noise ratio (PSNR) of the legitimate receiver is 4.14 dB higher than that of the eavesdropper.

Figures (13)

• Figure 1: Comparison between existing SemSteCom schemes and the proposed SemSteDiff scheme. (a) introduces the existing cover-edited SemSteCom schemes, which depend on embedding secret messages into cover images to obtain stego images. (b) introduces proposed coverless SemSteDiff scheme, which does not rely on cover images and generates stego images directly.
• Figure 2: The overview framework of SemSteDiff, where BLIP-based private key extractor obtains textual description of secret images as private keys, LLM-based public key generator produces public keys in pairs with private keys, conditional diffusion model-based coverless steganography module embeds keys' characteristic into attention mechanism to generate relative stego images, and JSCC-based semantic codec module achieves SemCom.
• Figure 3: Private and public key generation, where (a) shows that the private key is extracted from a secret image using BLIP model, (b) shows that public key is obtained by modifying private key under steganography requirement.
• Figure 4: Process of conditional diffusion model-based coverless steganography. First, the secret image encodes from pixel space into latent space by VAE encoder. Second, the secret latent vector adds noise under the guidance of private key using attention mechanism. Then, the public key guides the noisy vector to generate the stego latent vector. Finally, the VAE decoder decodes the latent vector into stego image.
• Figure 5: Performance of SemSteDiff scheme with DeepJSCC across four metrics. (a) shows the trend of PSNR under different trained SNRs and tested SNRs. (b), (c) and (d) show the trend of MSE, SSIM and LPIPS, respectively.

Theorems & Definitions (4)

• Remark 1
• Remark 2
• Remark 3
• Remark 4