Paper 2025/386 Original: eprint.iacr.org/2025/386 ↗

How Small Can S-boxes Be

Chenhao Jia
Tingting Cui
Qing Ling
Yan He
Kai Hu
Yu Sun
Meiqin Wang
Abstract

S-boxes are the most popular nonlinear building blocks used in symmetric-key primitives. Both cryptographic properties and implementation cost of an S-box are crucial for a good cipher design, especially for lightweight ones. This paper aims to determine the exact minimum area of optimal 4-bit S-boxes (whose differential uniform and linearity are both 4) under certain standard cell library. Firstly, we evaluate the upper and lower bounds upon the minimum area of S-boxes, by proposing a Prim-like greedy algorithm and utilizing properties of balanced Boolean functions to construct bijective S-boxes. Secondly, an SAT-aided automatic search tool is proposed that can simultaneously consider multiple cryptographic properties such as the uniform, linearity, algebraic degree, and the implementation costs such as area, and gate depth complexity. Thirdly, thanks to our tool, we manage to find the exact minimum area for different types of 4-bit S-boxes. The measurement in this paper uses the gate equivalent (GE) as standard unit under UMC 180 nm library, all 2/3/4-input logic gates are taken into consideration. Our results show that the minimum area of optimal 4-bit S-box is 11 GE and the depth is 3. If we do not use the 4-input gates, this minimum area increases to 12 GE and the depth in this case is 4, which is the same if we only use 2-input gates. If we further require that the S-boxes should not have fixed points, the minimum area continue increasing a bit to 12.33 GE while keeping the depth. Interestingly, the same results are also obtained for non-optimal 4-bit bijective S-boxes as long as their differential uniform $\mathcal{U}(S) <16$ and linearity $\mathcal{L}(S) < 8$ (i.e., there is no non-trivial linear structures) if only 2-input and 3-input gates are used. But the minimum area reduce to 9 GE if 4-input gates are involved. More strictly, if we require the algebraic degree of all coordinate functions of optimal S-boxes be 3, the minimum area is 14 GE with fixed point and 14.33 GE without fixed point, and the depth increases sharply to 8. Besides determining the exact minimum area, our tool is also useful to search for a better implementation of existing S-boxes. As a result, we find out an implementation of Keccak's 5-bit S-box with 17 GE. As a contrast, the designer's original circuit has an area of 23.33 GE, while the optimized result by Lu et al. achieves an area of 17.66 GE. Also, we find out the first optimized implementation of SKINNY's 8-bit S-box with 26.67 GE.

Metadata
Available format(s)
PDF
Category
Secret-key cryptography
Publication info
Preprint.
History
2025-02-28: last updated
2025-02-28: published
Short URL
https://ia.cr/2025/386
License
https://creativecommons.org/licenses/by/4.0/

BibTeX

@misc{cryptoeprint:2025/386,
      author = {Chenhao Jia and Tingting Cui and Qing Ling and Yan He and Kai Hu and Yu Sun and Meiqin Wang},
      title = {How Small Can S-boxes Be},
      howpublished = {Cryptology {ePrint} Archive, Paper 2025/386},
      year = {2025},
      url = {https://eprint.iacr.org/2025/386}
}
Note: In order to protect the privacy of readers, this archive does not use cookies or embedded third party content.