Almost perfect nonlinear (APN) functions are valued in symmetric cryptography for their optimal resistance to differential attacks. For maps on binary finite fields, the reduced polynomial representative is a standard way to write the function, but its formal derivative is not invariant under coordinate changes.

The new paper, posted on arXiv, notes that while differential uniformity does not depend on the chosen coordinates, the formal derivative of the reduced representative does. This discrepancy becomes important when lifting permutations from binary fields to Galois rings. The authors propose optimal reduced representatives for quadratic APN permutations in odd dimension, offering a canonical choice that avoids the coordinate ambiguity.

Because the abstract is brief, the full construction and proof are not detailed here. Still, the result appears to provide a cleaner algebraic framework for studying APN permutations and their lifts, which may be relevant for constructing cryptographic primitives over rings.