AbstractReversible Data Hiding (RDH) techniques based on Multiple Histograms Modification often struggle with optimal region classification, encryption robustness and memory efficiency. To address these challenges, this paper proposes a Reversible Data Hiding in Encrypted Images using Similarity-Navigated Graph Neural Networks with Efficient and Stable Thumbnail Preserving Encryption (RDH-EI-SNGNN-ESTPE). The Efficient and Stable Thumbnail Preserving Encryption (ESTPE) mechanism encrypts the image while preserving a low-resolution preview, which enable secure transmission without compromising usability. Then, Similarity-Navigated Graph Neural Networks (SNGNN) optimize embedding locations by efficiently navigating spatial similarities in the encrypted domain, ensuring adaptive and high-capacity data hiding. The encrypted and embedded image is safely transmitted with a thumbnail preview for verification. At the receiver end, the hidden data is extracted and the original image is fully reconstructed without loss. The experimental results on the COCO dataset exhibit a high Peak Signal-to-Noise Ratio of 45.8 dB, minimal memory usage of 92.7 MB and a high structural similarity of 0.987. The proposed method enhances adaptability and embedding capacity while maintaining encryption security.
TY - JOUR
AU - Bharathidasan, S.
AU - Selvi, V. Tamil
AU - Sathiya, V.
AU - Shajin, Francis H.
PY - 2025
DA - 2025/06/26
TI - Reversible data hiding in encrypted images using graph neural networks with thumbnail preserving encryption
JO - Signal, Image and Video Processing
SP - 802
VL - 19
IS - 10
AB - Reversible Data Hiding (RDH) techniques based on Multiple Histograms Modification often struggle with optimal region classification, encryption robustness and memory efficiency. To address these challenges, this paper proposes a Reversible Data Hiding in Encrypted Images using Similarity-Navigated Graph Neural Networks with Efficient and Stable Thumbnail Preserving Encryption (RDH-EI-SNGNN-ESTPE). The Efficient and Stable Thumbnail Preserving Encryption (ESTPE) mechanism encrypts the image while preserving a low-resolution preview, which enable secure transmission without compromising usability. Then, Similarity-Navigated Graph Neural Networks (SNGNN) optimize embedding locations by efficiently navigating spatial similarities in the encrypted domain, ensuring adaptive and high-capacity data hiding. The encrypted and embedded image is safely transmitted with a thumbnail preview for verification. At the receiver end, the hidden data is extracted and the original image is fully reconstructed without loss. The experimental results on the COCO dataset exhibit a high Peak Signal-to-Noise Ratio of 45.8 dB, minimal memory usage of 92.7 MB and a high structural similarity of 0.987. The proposed method enhances adaptability and embedding capacity while maintaining encryption security.
SN - 1863-1711
UR - https://doi.org/10.1007/s11760-025-04253-x
DO - 10.1007/s11760-025-04253-x
ID - Bharathidasan2025
ER -
|
50
[Full book]States of Immersion Across Media
