Existing deep learning-based reversible data hiding (RDH) predictors are affected by the difference of pixel complexity, which leads to the reduction of prediction accuracy. Therefore, this letter proposes an optimization framework tailored for RDH predictors, which integrates the local complexity of pixels into the predictor's regression optimization process. By analyzing the image's texture features, the framework adaptively determines the optimal prediction coefficients, thereby improving prediction accuracy. Notably, this optimization framework is versatile and can be applied to optimize other deep learning-based RDH predictors. Additionally, recognizing the critical role of interpolation strategies in RDH pixel prediction, we introduce a multi-scale fusion-enhanced interpolation network specifically designed for RDH, which integrates features across different scales to provide accurate reference pixels for subsequent predictions. Finally, experimental results demonstrate that the proposed method outperforms several advanced RDH predictors in terms of both prediction accuracy and embedding performance.

The proliferation of DeepFakes has heightened the necessity to distinguish between authentic and counterfeit faces. While numerous methods concentrate on detecting DeepFakes, only a few address safeguarding genuine faces from manipulation. This letter proposes a novel active forensics system for DeepFake forensics utilizing separable perceptual hash enhancement algorithm. A separable perceptual hash code specifically designed for face deep forgery is introduced, achieving robustness while maintaining sensitivity and imperceptibility when embedded within the original image. Additionally, a multi-scale perceptual smoothing loss function is employed to optimize perceptual similarity, structural smoothness, and embedding stability. As a result, this system ensures the consistence of confidential information both before and after manipulation, thereby enhancing the capability of face source detection and DeepFake identification. Experimental results demonstrate that the proposed scheme can effectively counter traditional deep forgery techniques while exhibiting significant potential in preserving personal privacy.

Aiming at the problem of most Robust Reversible Data Hiding (RRDH) schemes failing to anti geometric deformation attacks, a new RRDH algorithm based on Polar Harmonic Fourier Moments (PHFMs) is presented in this paper, thereby enhancing both the robustness of the embedded data and perceptual quality of the data-embedded image. Firstly, by leveraging the anti-geometric transformation and high-fidelity features of PHFMs, the image is transformed into its frequency domain for RRDH. Then, a quantitation index modulation (QIM) algorithm is designed to embed secret data into the integer part of PHFMs coefficients. By minimizing the differences between the secret-data-embedded image and the original image, the amount of compensation data is reduced. Meanwhile, a two-dimensional RDH scheme is further adopted to embed the compensation data, thus reducing the distortion of the full data-embedded image. Finally, the robustness of the embedded data and the fidelity of the full data-embedded image are both improved. The combination of PHFMs transformation and two-dimensional RDH enables the proposed RRDH algorithm to achieve high visual quality and strong resistance capability against geometric transformation attacks. Extensive experimental results demonstrate that the proposed RRDH algorithm outperforms other state-of-the-art techniques.