Home / Cryptography & Math / New Research Breakthrough Sheds Light on Quadratic APN Functions in Dimension Ten

New Research Breakthrough Sheds Light on Quadratic APN Functions in Dimension Ten

A major new study has advanced the mathematical understanding of cryptographic foundations, focusing on the complex structures of Almost Perfect Nonlinear (APN) functions. Released as Paper 2026/2889 on the Cryptology ePrint Archive, the research was conducted by Emanuele Bellini of the Technology Innovation Institute. The work deepens the theoretical framework surrounding APN functions over finite fields, offering fresh insights into high-dimensional cryptographic components, uncovering new mathematical classes, and resolving nearly all prime-order self-equivalence types in dimension ten.

The investigation centers on APN functions of the vector space over the finite field of two elements, denoted as $mathbbF_2^n$. In the realm of cryptography and coding theory, these functions hold a position of paramount importance. They achieve the lowest possible differential uniformity, making them exceptionally resilient against differential cryptanalysis—a powerful technique used by cryptanalysts to break block ciphers and other symmetric encryption algorithms. Because security engineers rely heavily on substitution boxes (S-boxes) built upon functions with optimal differential properties, mapping out the landscape of APN functions is crucial for designing secure, future-proof cryptographic systems.

Despite their vital role in theoretical and applied cryptography, understanding the full classification and behavior of APN functions remains a notoriously difficult mathematical challenge. The behavior of these functions changes drastically depending on whether the dimension $n$ is even or odd. For odd dimensions, numerous APN permutations are known to exist. However, the even-dimensional case presents a starkly different and much more elusive landscape.

To date, the only known APN permutation in an even dimension exists in dimension six

. This specific permutation is fundamentally equivalent to a quadratic APN function characterized by 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 captivated researchers for decades.

As dimensions scale upward, the computational complexity of analyzing these functions explodes. Previous efforts to catalog quadratic APN classes in dimension eight revealed that out of roughly 3.8 million known classes, a tiny fraction—at most 0.34 percent—possess a nontrivial linear self-equivalence. By contrast, when moving to dimension ten, a shift occurs: all 25 literature classes considered in current research share this characteristic. Building upon this foundation, Bellini’s new study focuses on the comprehensive analysis of all quadratic APN functions of $mathbbF_2^10$ that feature a self-equivalence of prime order.

When a quadratic APN function possesses a nontrivial linear self-equivalence, it is typically constrained to a much smaller linear subspace, allowing mathematicians to analyze its underlying structure more effectively. Up to conjugacy and power similarity, self-equivalences of prime order manifest in 299 distinct types within dimension ten. The core achievement of the new paper is the systematic resolution of 298 of these 299 types.

To accomplish this monumental task, the research employed a combination of theoretical tools and heavy computational power. Elementary lemmas were successfully deployed to exclude 280 of the types outright. To ensure absolute mathematical rigor, exhaustive searches were subsequently conducted, confirming the exclusion of 242 of those types through direct computational verification. The remaining 18 types required even more intensive computational methods, ultimately being settled through exhaustive searches backed by exact equivalence certificates.

Through this rigorous classification process, up to extended affine equivalence, the 298 resolved types were found to contain 29 distinct classes of quadratic APN functions. Significantly, four of these classes are entirely new to mathematics and cryptography, expanding the known catalog of high-dimensional APN structures.

The research also sheds light on how these quadratic APN functions relate to coefficients situated in specific subfields, an area of significant interest for algebraic constructions. The findings reveal that exactly 24 of the 29 classes contain a function with coefficients in $mathbbF4$. Furthermore, all three classes that contain a function with coefficients in $mathbbF32$ are also found neatly within this group of 24, highlighting a structured hierarchy among subfield coefficients.

Despite the sweeping success in resolving 298 out of the 299 prime-order self-equivalence types, a single enigmatic type remains open. This remaining type consists of the quadratic canonical triplicate functions. Mathematically, these specific functions correspond to maximum additive Hermitian rank-distance codes, representing a unique geometric and algebraic structure that continues to resist complete classification.

To probe this final frontier, three different independent computational searches were deployed. However, these searches consistently yielded only the invariants of 16 classes, all of which are already securely cataloged among the 29 discovered classes. Based on this extensive computational evidence, the author conjectures that no other functions exist within this open type, pointing toward a closed and complete system.

Another critical finding of the research concerns the relationship between these functions and permutations. In the study of block ciphers, finding APN functions that are equivalent to permutations is of the highest practical value. However, the analysis delivered a definitive negative result regarding the settled types: none of the quadratic APN functions contained within the 298 settled types are Carlet-Charpin-Zinoviev equivalent to a permutation. This reinforces the ongoing mystery surrounding the scarcity of even-dimensional APN permutations and underscores the unique constraints governing functions in dimension ten.

By mapping out the intricate taxonomy of quadratic APN functions with prime-order self-equivalences, this research provides the cryptographic community with a clearer, more rigorous map of the algebraic terrain in dimension ten. As theoretical cryptographers continue to push the boundaries of what is mathematically possible in even dimensions, studies of this scale lay the essential groundwork for the next generation of cryptographic design and analysis.

Tagged: