In a significant development for cryptographic research and the mathematical foundations of coding theory, researcher Emanuele Bellini of the Technology Innovation Institute has released a comprehensive new study detailing the properties of quadratic Almost Perfect Nonlinear (APN) functions of $mathbbF_2^10$. Published as Paper 2026/2889 on the Cryptology ePrint Archive, the work explores the intricate landscape of these functions when they possess a self-equivalence of prime order, successfully classifying a vast majority of types and uncovering four entirely new classes of quadratic APN functions.
The research dives deep into the complex world of vector spaces over finite fields, specifically focusing on functions that offer the lowest possible differential uniformity. In cryptography, particularly in the design of symmetric ciphers and substitution boxes, or S-boxes, functions with optimal resistance against differential cryptanalysis are of paramount importance. APN functions represent the gold standard in this regard, minimizing the vulnerability of cryptographic algorithms to sophisticated attacks by ensuring that input differences yield output differences as uniformly as possible.
Despite decades of intensive study by mathematicians and cryptographers, the behavior and classification of APN functions in even dimensions remain notoriously difficult to pin down. The broader mathematical landscape features a persistent puzzle regarding APN permutations—functions that are both APN and bijective. To date, the only known APN permutation in an even dimension exists in dimension six, where it is uniquely equivalent to a quadratic APN function featuring a nontrivial linear self-equivalence. For even dimensions greater than or equal to eight, no APN permutations are currently known to exist, a frustrating barrier that has driven researchers to examine the fine-grained structural properties of quadratic APN functions instead.
To understand the scale of the challenge tackled in the new paper, one must look at what is known about dimensions eight and ten. In dimension eight, a previously published and widely cited collection catalogs approximately 3.8 million quadratic APN classes. Interestingly, within that massive pool, a mere 0.34 percent possess a nontrivial linear self-equivalence, demonstrating that such symmetry is a rare and restrictive feature. Dimension ten presents a contrasting baseline: all 25 literature classes considered in Bellini’s study share this characteristic, possessing a nontrivial linear self-equivalence.
It is this specific subset of quadratic APN functions in $mathbbF_2^10$—those possessing a self-equivalence of prime order—that forms the primary focus of the new research. When a function exhibits a self-equivalence of this nature, it is typically constrained, confining the function to a much smaller linear space and rendering its internal structure more amenable to rigorous analysis, classification, and computational exhaustion.
According to the study, when evaluated up to conjugacy and power similarity, self-equivalences of prime order reveal a staggering variety, spanning 299 distinct types. Bellini’s work systematically addresses this entire landscape, successfully settling 298 of these types. The methodology relies on a combination of theoretical tools and heavy computational power. Elementary lemmas were deployed to successfully exclude 280 types, a finding that was further confirmed through exhaustive computational searches for 242 of them. The remaining 18 types, which resisted simpler exclusion techniques, were ultimately settled through advanced exhaustive searches utilizing exact equivalence certificates, ensuring the mathematical rigor and correctness of the classification.
The implications of settling these 298 types extend further when examined through the lens of extended affine equivalence. Within these settled types lie 29 distinct classes of quadratic APN functions. Crucially, four of these classes are entirely new to mathematical literature, expanding humanity’s formal catalog of these optimal cryptographic mappings.
This classification effort also sheds light on the behavior of quadratic APN functions whose coefficients lie within specific subfields. The research reveals precise structural patterns regarding subfield coefficients. Exactly 24 of the 29 classes contain a function with coefficients in $mathbbF4$. Furthermore, the study notes a striking inclusion hierarchy: all three classes that contain a function with coefficients in $mathbbF32$ are also found within those 24 classes, highlighting a neat nested relationship among these algebraic structures.
Despite the breadth of the success achieved across 298 of the prime-order types, one type remains conspicuously open. This remaining open type consists of the quadratic canonical triplicate functions. These specific functions hold deep theoretical significance because they correspond directly to maximum additive Hermitian rank-distance codes, connecting the study of APN functions to advanced coding theory and geometry.
To probe this final frontier, Bellini deployed three different computational searches. However, these searches yielded only the invariants of 16 classes, all of which are already accounted for among the 29 established classes. Based on these exhaustive computational efforts, the author posits a compelling conjecture: there are likely no other functions to be found within this open type.
The study also addresses a long-standing question regarding the relationship between quadratic APN functions and permutations. In the quest for even-dimensional APN permutations—which, as noted, remain elusive beyond dimension six—researchers frequently check whether known functions are Carlet-Charpin-Zinoviev (CCZ) equivalent to a permutation. Bellini’s findings close that door for this specific family, confirming that no quadratic APN function residing within the 298 settled types is CCZ-equivalent to a permutation.
By pushing the boundaries of what is computationally and theoretically manageable in dimension ten, this work provides the cryptographic and mathematical communities with a firmer footing, refining the taxonomy of APN functions and illuminating the structural boundaries that continue to define modern finite field theory.

