Differential privacy comes in several flavors, and the choice among them can change both the privacy analysis and the resulting utility. A new arXiv preprint proposes a unified information-theoretic framework that covers four mainstream curve-based privacy notions, aiming to bring them under one accounting model.
The abstract does not name the four notions or give technical details, but it signals that the unification is built on information-theoretic principles. This suggests the authors are looking for a common way to measure privacy loss and information loss across different definitions, rather than treating each as a separate case.
If successful, such a framework could make it easier to compare results across DP variants and to reason about trade-offs between privacy and utility in a more principled way. As with any preprint, the claims have not yet been peer-reviewed, and the full method remains to be examined.