Substitution-Permutation Networks, Pseudorandom Functions, and Natural Proofs

Conference paper

DOI: 10.1007/978-3-642-32009-5_5

Volume 7417 of the book series Lecture Notes in Computer Science (LNCS)
Cite this paper as:
Miles E., Viola E. (2012) Substitution-Permutation Networks, Pseudorandom Functions, and Natural Proofs. In: Safavi-Naini R., Canetti R. (eds) Advances in Cryptology – CRYPTO 2012. Lecture Notes in Computer Science, vol 7417. Springer, Berlin, Heidelberg


This paper takes a new step towards closing the troubling gap between pseudorandom functions (PRF) and their popular, bounded-input-length counterparts. This gap is both quantitative, because these counterparts are more efficient than PRF in various ways, and methodological, because these counterparts usually fit in the substitution-permutation network paradigm (SPN) which has not been used to construct PRF.

We give several candidate PRF \(\mathcal {F}_i\) that are inspired by the SPN paradigm. This paradigm involves a “substitution function” (S-box). Our main candidates are:

\(\mathcal {F}_1 : \{0, 1\}^n \rightarrow \{0, 1\}^n\) is an SPN whose S-box is a random function on b bits given as part of the seed. We prove unconditionally that \(\mathcal {F}_1\) resists attacks that run in time \(\le 2^{\epsilon b}\). Setting \(b = \omega (\lg n)\) we obtain an inefficient PRF, which however seems to be the first such construction using the SPN paradigm.

\(\mathcal {F}_2 : \{0, 1\}^n \rightarrow \{0, 1\}^n\) is an SPN where the S-box is (patched) field inversion, a common choice in practical constructions. \(\mathcal {F}_2\) is computable with Boolean circuits of size \(n \cdot \log ^{O(1)} n\), and in particular with seed length \(n \cdot \log ^{O(1)} n\). We prove that this candidate has exponential security \(2^{\Omega (n)}\) against linear and differential cryptanalysis.

\(\mathcal {F}_3 : \{0, 1\}^n \rightarrow \{0, 1\}\) is a non-standard variant on the SPN paradigm, where “states” grow in length. \(\mathcal {F}_3\) is computable with size \(n^{1+\epsilon }\), for any \(\epsilon > 0\), in the restricted circuit class \(\mathrm {TC}^0\) of unbounded fan-in majority circuits of constant-depth. We prove that \(\mathcal {F}_3\) is almost 3-wise independent.

\(\mathcal {F}_4 : \{0, 1\}^n \rightarrow \{0, 1\}\) uses an extreme setting of the SPN parameters (one round, one S-box, no diffusion matrix). The S-box is again (patched) field inversion. We prove that this candidate fools all parity tests that look at \(\le 2^{0.9n}\) outputs.

Assuming the security of our candidates, our work also narrows the gap between the “Natural Proofs barrier” [Razborov & Rudich; JCSS ’97] and existing lower bounds, in three models: unbounded-depth circuits, \(\mathrm {TC}^0\) circuits, and Turing machines. In particular, the efficiency of the circuits computing \(\mathcal {F}_3\) is related to a result by Allender and Koucky [JACM ’10] who show that a lower bound for such circuits would imply a lower bound for \(\mathrm {TC}^0\).

Download to read the full conference paper text

Copyright information

© International Association for Cryptologic Research 2012 2012

Authors and Affiliations

  1. 1.Northeastern UniversityBostonUSA