Research map / Image generation / Autoregressive and latent generation
Hiding Images in Plain Sight: Deep Steganography
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Adding the secret-image reconstruction error term with weight beta (vs. cover-only encoding) | Deep-Stego hiding network · ImageNet test (1,000 pairs) | SSE per pixel per channel: cover error rises from 0.1 (beta=0) to 2.8-6.4 across beta values; secret error 2.8-3.6 | Mixed |
| Increasing beta from 1.00 to 1.25 (ablation) | Deep-Stego · ImageNet test (1,000 pairs) | cover SSE per pixel per channel: cover 3.0 -> 6.4 while secret 3.2 -> 2.8 | Hurt |
| Replacing with simple 4-LSB substitution instead of learned hiding | LSB baseline · ImageNet test | avg pixel error per channel: cover 5.4 and secret ~4.0 vs Deep-Stego 2.8-3.0 cover | Hurt |
| Adding a residual-correlation penalty minimizing corr(Rc, S) (ablation) | Deep-Stego · Residual analysis with cover leaked (qualitative) | secret features visible in residual | Helped |
| Adding the residual-correlation penalty (ablation) | Deep-Stego · Reconstruction quality | container pixel error | Hurt |
| Using a learned network (Deep-Stego) instead of LSB substitution against steganalysis | Deep-Stego containers · StegExpose and CNN steganalysis | detection rate: StegExpose near random; CNN 90-95% vs LSB >99% | Helped |
Closest papers by meaning
- Neural Cover Selection for Image Steganography (NeurIPS 2024 · similarity 0.86)
- Purified and Unified Steganographic Network (CVPR 2024 · similarity 0.86)
- Robust Message Embedding via Attention Flow-Based Steganography (CVPR 2025 · similarity 0.85)
- Efficient and Separate Authentication Image Steganography Network (ICML 2025 · similarity 0.85)
- StegoGAN: Leveraging Steganography for Non-Bijective Image-to-Image Translation (CVPR 2024 · similarity 0.85)
- Hiding Images in Diffusion Models by Editing Learned Score Functions (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.