发表机构
UCSB; UCLA(加州大学圣塔芭芭拉分校; 加州大学洛杉矶分校)
机构由 AI 辅助整理,请以论文原文为准。AI 中文总结
本文在经典通用群模型中,利用素数阶循环群无条件构造了针对NP问题的见证加密方案,并首次证明了齐次MinRank问题的超常数因子NP困难性近似结果。
AI 中文摘要
我们在经典通用群模型中,使用一个普通的素数阶循环群,无条件地构造了针对NP问题的见证加密方案。对于规模为$n$的SAT实例,加密算法在poly$(n)$时间内运行,任何满足性赋值可用于在poly$(n)$时间内解密,正确性错误为$2^{-n^{\Omega(1)}}$。如果不存在满足性赋值,那么每个进行至多$n^{\Theta(\log n)}$次群查询的通用敌手,其区分优势至多为$n^{-\Theta(\log n)}$。在此过程中,我们首次证明了在随机化多项式时间归约下,齐次MinRank问题的近似难度具有超常数因子的NP困难性,即使在秩一见证具有布尔右因子时,也能实现对数级别的间隙。
英文摘要
We unconditionally construct witness encryption for NP in the classical generic-group model, using an ordinary cyclic group of prime order. For SAT instances of size $n$, the encryption algorithm runs in time poly$(n)$, and any satisfying assignment can be used to decrypt in poly$(n)$ time with correctness error $2^{-n^{Ω(1)}}$. If no satisfying assignment exists, then every generic adversary making at most $n^{Θ(\log n)}$ group queries has distinguishing advantage at most $n^{-Θ(\log n)}$. Along the way, we prove the first superconstant-factor NP-hardness of approximation result for homogeneous MinRank under randomized polynomial-time reductions, achieving a logarithmic gap even when the rank-one witness has a Boolean right factor.