International Association
for Cryptologic Research

A central advantage of deploying cryptosystems is that the security of large high-sensitive data sets can be reduced to the security of a very small key. The most popular way to manage keys is to use a $(t,n)-$threshold secret sharing scheme: a user splits her/his key into $n$ shares, distributes them among $n$ key servers, and can recover the key with the aid of any $t$ of them. However, it is vulnerable to device destruction: if all key servers and user's devices break down, the key will be permanently lost. We propose a $\mathrm{\underline{D}}$estruction-$\mathrm{\underline{R}}$esistant $\mathrm{\underline{K}}$ey $\mathrm{\underline{M}}$anagement scheme, dubbed DRKM, which ensures the key availability even if destruction occurs. In DRKM, a user utilizes her/his $n^{*}$ personal identification factors (PIFs) to derive a cryptographic key but can retrieve the key using any $t^{*}$ of the $n^{*}$ PIFs. As most PIFs can be retrieved by the user $\textit{per se}$ without requiring $\textit{stateful}$ devices, destruction resistance is achieved. With the integration of a $(t,n)-$threshold secret sharing scheme, DRKM also provides $\textit{portable}$ key access for the user (with the aid of any $t$ of $n$ key servers) before destruction occurs. DRKM can be utilized to construct a destruction-resistant cryptosystem (DRC) in tandem with any backup system. We formally prove the security of DRKM, implement a DRKM prototype, and conduct a comprehensive performance evaluation to demonstrate its high efficiency. We further utilize Cramer's Rule to reduce the required buffer to retrieve a key from 25 MB to 40 KB (for 256-bit security).