Content area
Reversible data hiding in encrypted images (RDH-EI) integrates encryption with information hiding, enabling the embedding of additional data while ensuring full recovery of the original image, widely used in multimedia data protection and forensics. However, with the increasingly serious problem of data islands, the existing RDH-EI schemes for end-to-end communication scenarios can no longer meet the application requirements for multi-party data sharing. For solving this problem, this paper proposes an RDH-EI method based on Lasso regression predictor and dynamic secret sharing. First, Lasso regression predictor is proposed, which avoids the problem of over-fitting by regularizing the loss function. Then, to reserve more embeddable room, we map the prediction error and use arithmetic coding for compression. Next, the original image and auxiliary information are shared through dynamic secret sharing algorithm, and they can be transmitted to the corresponding data hiders respectively. According to the reserved room, each data hider allows for the discrete implantation of the secret bits into the corresponding encrypted image. The receiver can successfully extract the secret bits and recreate the cover image without distortion by gathering a specific percentage of marked images. Experimental results demonstrate that the suggested Lasso regression predictor has a greater prediction accuracy, outperforming contemporary methods for embedding rate. Additionally, it has great security and strong key sensitivity.