In a significant development for cryptographic research and theoretical computer science, a team of researchers has announced a breakthrough construction of Boolean functions on 375 variables that successfully exceeds the classic Patterson-Wiedemann direct-sum bound. The research, conducted by Deng Tang from Shanghai Jiao Tong University and Subhamoy Maitra from the Indian Statistical Institute, marks a notable advance in the study of cryptographic properties such as nonlinearity, which plays a critical role in the security and resilience of symmetric ciphers against cryptographic attacks.
The new findings, detailed in a recent paper published on the Cryptology ePrint Archive as Paper 2026/2680, address a foundational challenge in coding theory and cryptography. Specifically, the authors have demonstrated the existence of both general and balanced Boolean functions on 375 variables whose nonlinearity surpasses the theoretical limits previously established through the direct-sum combination of Patterson-Wiedemann functions and bent functions. This achievement pushes past mathematical boundaries that have guided the field for decades.
To understand the significance of the new research, it is helpful to look back at the historical framework established by Wayne Patterson and Douglas Wiedemann in their seminal 1983 work published in the IEEE Transactions on Information Technology. Patterson and Wiedemann constructed Boolean functions on 15 variables with a nonlinearity of 16,276, a figure that was strictly greater than the prevailing bent concatenation bound of that era. By taking the direct sum of these 15-variable functions with a bent function, researchers could generate higher-variable functions yielding a specific asymptotic nonlinearity for every odd number of variables greater than or equal to 15.
Decades later, in 2019, researcher Kai-Uwe Schmidt further refined the theoretical landscape by proving in the Journal of Combinatorial Theory, Series A, that the asymptotic maximum nonlinearity for such functions approaches a specific upper limit as the number of variables increases to infinity. For many years, the direct-sum method combining a 15-variable Patterson-Wiedemann function with a 360-variable bent function served as a powerful benchmark, yielding a high degree of nonlinearity for 375 variables. However, the new work by Tang and Maitra proves that this long-standing direct-sum bound is not the ultimate ceiling.
In their paper, Tang and Maitra construct Boolean functions on 375 variables with a nonlinearity that exceeds the traditional benchmark by a decisive margin. To be precise, the direct sum of a 15-variable Patterson-Wiedemann function and a 360-variable bent function yields a nonlinearity defined mathematically as $2^374-2^187+20cdot2^180$, which simplifies to $2^374-2^187+81920cdot2^168$. In contrast, the new construction achieves a nonlinearity greater than $2^374-442367cdot2^168$, or $2^374-2^187+81921cdot2^168$. This means the new construction improves upon the traditional direct-sum value by more than $2^168$, a clear and quantifiable leap beyond prior limitations.
The methodology behind this breakthrough relies on sophisticated algebraic techniques involving finite fields, Gauss sums, and linear feasibility constraints. The authors begin by letting $H$ be the multiplicative subgroup of order $(2^375-1)/751$ within the finite field extension $mathbbF_2^375^*$. In this framework, the support of the newly constructed Boolean function $f$ is designed as the disjoint union of 375 prescribed cosets combined with a specific subset $B$ contained within the subgroup $H$.
A critical component of the proof involves the evaluation of Gauss sums, which effectively reduces the complex contribution of the 375 prescribed cosets down to a manageable computation involving 751 integers. Once this reduction is achieved, the conditions required for the subset $B$ translate into a zero-one linear feasibility problem. By employing a potential criterion for simultaneous linear constraints, Tang and Maitra are able to rigorously prove the existence of a single subset $B$ within the subgroup $H$ that satisfies all required constraints simultaneously across every frequency.
Furthermore, the researchers show that this foundational result can be extended broadly. By taking direct sums of their newly constructed functions with standard bent functions, the approach successfully extends the improved nonlinearity bounds to every odd number of variables greater than or equal to 375. This structural adaptability ensures that the implications of the work stretch far beyond the specific case of 375 variables, providing a generalized pathway for constructing high-nonlinearity functions in higher dimensions.
In addition to addressing general Boolean functions, the authors also tackle the crucial requirement of balancedness, which is essential for preventing statistical biases in cryptographic applications such as stream ciphers and block ciphers. The initial support construction developed by Tang and Maitra does not inherently require balancedness. However, as a natural and vital follow-up, the researchers successfully derived balanced functions sharing the exact same strict lower bound. They achieved this by altering the minimum possible number of values using a statistical technique known as sampling without replacement.
Significantly, both the primary subsets $B$ and the specific sets of modified points required to achieve balancedness are established through rigorous existence proofs. These mathematical guarantees demonstrate that such optimal structures do indeed exist within the vast space of Boolean functions, even if explicitly computing every single element presents formidable computational challenges.
The publication of Paper 2026/2680 on the Cryptology ePrint Archive adds an important chapter to the ongoing exploration of cryptographic mathematics. By breaking past the Patterson-Wiedemann direct-sum bound for 375 variables and offering a robust framework for generating balanced functions with equivalent properties, Tang and Maitra have expanded the theoretical boundaries available to cryptographers and mathematicians designing the next generation of secure communication systems.






