Downward recompression robust JPEG steganography via efficient content-adaptive embedding

Downward recompression robust JPEG steganography via efficient content-adaptive embedding

연구 분야: Strategies

학회: Multimedia Tools and Applications

The widespread sharing of JPEG images over Social Networking Platforms (SNPs) presents challenges for steganography due to aggressive re-compression applied by these platforms, particularly in the downward robust (DR) scenario where high-quality images are compressed to lower quality factors. Existing robust steganography methods often overlooks the DR challenges and relies on computationally expensive methods requiring capacity-consuming error-correcting codes (ECCs). To address this gap, this paper presents CEDAR, an Efficient Content-adaptive Downward Robust Algorithm for JPEG steganography. CEDAR uniquely achieves robustness without ECCs by strategically identifying resilient DCT coefficient locations based purely on analyzing cover and anticipated channel quantization tables, coupled with an efficient adaptive embedding strategy that modifies coefficient parity with minimal distortion. Comprehensive experiments, comparing CEDAR against state-of-the-art methods including DMMR, GMAS and SS, demonstrate its superior performance across benchmark images under various payloads and downward re-compression scenarios. CEDAR sustains a low message extraction error rate, higher undetectability against statistical attacks, imperceptible visual quality, and significantly lower computational complexity. CEDAR offers a practical, efficient, and secure solution for robust steganography over modern SNPs.