EUVIMEDEUVIMED entdecken
Crossref

AAMC-RL: Reinforcement Learning-Based Automatic Asymmetric Cost Learning Steganography via Multi-Steganalyzer Consensus

Huiyan Chang, Dacheng Zhou, Hongchao Hu, Tao Hu, Quan Ren

IEEE Signal Processing Letters · 2026 · 33 · 1576-1580

DOI: 10.1109/lsp.2026.3682190 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Huiyan Chang und 4 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Gefördert von: Major Science and Technology Special Projects of Henan Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The asymmetric cost learning-based image steganography improves steganographic security by comprehensively considering the statistical dependencies among adjacent pixels. However, existing approaches still face challenges in network architecture design and reward signal formulation. To address these limitations, this paper proposes a novel asymmetric cost learning steganography framework (AAMC-RL). The proposed framework employs a dual-branch policy network integrated with a Gated Bidirectional Cross-Attention mechanism (GBCA), effectively mitigating the constraints of symmetric cost and alleviating training imbalance across individual branches. Furthermore, AAMC-RL incorporates multiple steganalyzers to construct an ensemble environment network and introduces a dual reward mechanism combining gradient consensus among steganalyzers and content consistency. Extensive experimental and ablation studies demonstrate that the proposed architecture achieves superior security performance to current state-of-the-art methods in asymmetric cost learning-based steganography.

Förderung

  • Major Science and Technology Special Projects of Henan Province

Zitieren

APA

Chang, H., Zhou, D., Hu, H., Hu, T., & Ren, Q. (2026). AAMC-RL: Reinforcement Learning-Based Automatic Asymmetric Cost Learning Steganography via Multi-Steganalyzer Consensus. IEEE Signal Processing Letters, 33, 1576-1580. https://doi.org/10.1109/lsp.2026.3682190

Vancouver

Chang H, Zhou D, Hu H, Hu T, Ren Q. AAMC-RL: Reinforcement Learning-Based Automatic Asymmetric Cost Learning Steganography via Multi-Steganalyzer Consensus. IEEE Signal Processing Letters. 2026;33:1576-1580. doi:10.1109/lsp.2026.3682190

BibTeX
@article{Chang2026,
  author = {Chang, Huiyan and Zhou, Dacheng and Hu, Hongchao and Hu, Tao and Ren, Quan},
  title = {AAMC-RL: Reinforcement Learning-Based Automatic Asymmetric Cost Learning Steganography via Multi-Steganalyzer Consensus},
  journal = {IEEE Signal Processing Letters},
  year = {2026},
  volume = {33},
  pages = {1576--1580},
  doi = {10.1109/lsp.2026.3682190}
}
Kostenlos starten →