Paper 2023/289

Lower Bounds for Secret-Sharing Schemes for k-Hypergraphs

Abstract

A secret-sharing scheme enables a dealer, holding a secret string, to distribute shares to parties such that only pre-defined authorized subsets of parties can reconstruct the secret. The collection of authorized sets is called an access structure. There is a huge gap between the best known upper bounds on the share size of a secret-sharing scheme realizing an arbitrary access structure and the best known lower bounds on the size of these shares. For an arbitrary $n$-party access structure, the best known upper bound on the share size is $2^{O(n)}$. On the other hand, the best known lower bound on the total share size is much smaller, i.e., $\Omega(n^2/\log (n))$ [Csirmaz, \emph{Studia Sci. Math. Hungar.}]. This lower bound was proved more than 25 years ago and no major progress has been made since. In this paper, we study secret-sharing schemes for $k$-hypergraphs, i.e., for access structures where all minimal authorized sets are of size exactly $k$ (however, unauthorized sets can be larger). We consider the case where $k$ is small, i.e., constant or at most $\log (n)$. The trivial upper bound for these access structures is $O(n\cdot \binom{n-1}{k-1})$ and this can be slightly improved. If there were efficient secret-sharing schemes for such $k$-hypergraphs (e.g., $2$-hypergraphs or $3$-hypergraphs), then we would be able to construct secret-sharing schemes for arbitrary access structures that are better than the best known schemes. Thus, understanding the share size required for $k$-hypergraphs is important. Prior to our work, the best known lower bound for these access structures was $\Omega(n \log (n))$, which holds already for graphs (i.e., $2$-hypergraphs). We improve this lower bound, proving a lower bound of $\Omega(n^{2-1/(k-1)}/k)$ on the total share size for some explicit $k$-hypergraphs, where $3 \leq k \leq \log (n)$. For example, for $3$-hypergraphs we prove a lower bound of $\Omega(n^{3/2})$. For $\log (n)$-hypergraphs, we prove a lower bound of $\Omega(n^{2}/\log (n))$, i.e., we show that the lower bound of Csirmaz holds already when all minimal authorized sets are of size $\log (n)$. Our proof is simple and shows that the lower bound of Csirmaz holds for a simple variant of the access structure considered by Csirmaz. Using our results, we prove a near quadratic separation between the required share size for realizing an explicit access structure and the monotone circuit size describing the access structure,i.e., the share size in $\Omega(n^2/\log(n))$ and the monotone circuit size is $O(n\log(n))$ (where the circuit has depth $3$).