There are advantages and disadvantages to both robust and cryptographic hash methods. Integrating the qualities of robustness and cryptographic confidentiality would be highly desirable. However, the challenge is that the concept of similarity is not applicable to cryptographic hashes, preventing direct comparison between robust and cryptographic hashes. Therefore, when incorporating robust hashes into cryptographic hashes, it becomes essential to develop methods that effectively capture the intrinsic properties of robust hashes without compromising their robustness. In order to accomplish this, it is necessary to anticipate the hash bits that are most susceptible to modification, such as those that are affected by JPEG compression. Our work demonstrates that the prediction accuracy of existing approaches can be significantly improved by using a new hybrid hash comparison strategy.
Niklas Bunzel, Martin Steinebach, Marius Leon Hammann, Huajian Liu, "Prediction of Flipped Bits in Robust Image Hashes by Machine Learning" in Electronic Imaging, 2024, pp 339-1 - 339-8, https://doi.org/10.2352/EI.2024.36.4.MWSF-339