AIIC AI Intelligence Centre

SOURCE-LINKED INTELLIGENCE

Style as Cover: Deep Image Steganography via Stylized Transmission

arXiv · AI, language, vision and robotics · article · Sep 18, 2026 · UTC

Image steganography hides secret message within normal images, with most existing works relying on cover-preserving transmission. However, such a paradigm becomes vulnerable once the original cover is exposed or can be reliably approximated. In this paper, we propose StyleStegaNet, a stylized image hiding framework that replaces cover matching with style-concealment transmission. Instead of transmitting a cover-like stego image, StyleStegaNet generates stylized stego images conditioned on publicly available style references, redefining steganography invisibility from cover-preserving concealme

Read original source ↗ Open in workspace

recordType
paper
region
Global

Evidence & attribution

First collected: 2026-09-23T14:01:59.594Z. This is not the publication date.