Reversible Data Hiding in Encrypted Images using Local Difference of Neighboring Pixels

This paper presents a reversible data hiding in encrypted image (RDHEI), which divides image into non-overlapping blocks. In each block, central pixel of the block is considered as leader pixel and others as follower ones. The prediction errors between the intensity of follower pixels and leader ones are calculated and analyzed to determine a feature for block embedding capacity. This feature indicates the amount of data that can be embedded in a block. Using this pre-process for whole blocks, we vacate rooms before the encryption of the original image to achieve high embedding capacity. Also, using the features of all blocks, embedded data is extracted and the original image is perfectly reconstructed at the decoding phase. In effect, comparing to existent RDHEI algorithms, embedding capacity is significantly increased in the proposed algorithm. Experimental results confirm that the proposed algorithm outperforms state of the art ones.

💡 Research Summary

The paper proposes a novel reversible data hiding in encrypted images (RDHEI) scheme that leverages the local difference between neighboring pixels within non‑overlapping blocks. Each block is divided into a central “leader” pixel and surrounding “follower” pixels. By computing the prediction error (the intensity difference) between each follower and its leader, the authors obtain a statistical profile of the block. This profile is used to derive a Block Capacity Feature (BCF), which quantifies how many bits can be safely embedded in that block without compromising reversibility.

The workflow consists of three main stages. First, the original image is partitioned into blocks, and a content‑owner key (α₁) permutes the block order. Then, for each permuted block, the BCF is calculated based on the distribution of prediction errors. After BCF extraction, the permuted image is encrypted by a stream‑cipher XOR operation using a second secret key (α₂). This double‑layer encryption (permutation + XOR) ensures that both pixel values and their local differences are concealed.

In the embedding phase, the data hider first embeds the BCFs themselves into the encrypted image, followed by the actual secret payload. The embedding follows a hierarchical procedure: higher‑level BCF bits are placed first, providing a map that later guides the extraction of the payload. An optional third key (α₃) can be used to encrypt the BCFs before insertion, preventing an attacker from learning the embedding capacity distribution.

During extraction, the receiver (who may be either the content owner or the data hider) reverses the XOR with α₂ and restores the original block order using α₁. The BCFs are then extracted hierarchically; their values directly indicate where and how many payload bits are stored. Finally, using the leader‑follower relationship and the known BCF, the original pixel intensities are perfectly reconstructed without requiring the data‑hider’s key. This separable property means that the content owner can recover the pristine image solely with his own keys, while the data hider can retrieve the hidden message independently.

The authors compare their method against state‑of‑the‑art RDHEI schemes, especially those based on the high‑capacity approach of reference