Multi-Client Functional Encryption for Separable Functions

In this work, we provide a compiler that transforms a single-input functional encryption scheme for the class of polynomially bounded circuits into a multi-client functional encryption (MCFE) scheme for the class of separable functions. An n-input function f is called separable if it can be described as a list of polynomially bounded circuits \(f^1, \dots , f^n\) s.t. \(f(x_1, \dots , x_n)= f^1(x_1)+ \dots + f^n(x_n)\) for all \(x_1,\dots , x_n\). Our compiler extends the works of Brakerski et al. [Eurocrypt 2016] and of Komargodski et al. [Eurocrypt 2017] in which a generic compiler is proposed to obtain multi-input functional encryption (MIFE) from single-input functional encryption. Our construction achieves the stronger notion of MCFE but for the less generic class of separable functions. Prior to our work, a long line of results has been proposed in the setting of MCFE for the inner-product functionality, which is a special case of a separable function. We also propose a modified version of the notion of decentralized MCFE introduced by Chotard et al. [Asiacrypt 2018] that we call outsourceable mulit-client functional encryption (OMCFE). Intuitively, the notion of OMCFE makes it possible to distribute the load of the decryption procedure among at most n different entities, which will return decryption shares that can be combined (e.g., additively) thus obtaining the output of the computation. This notion is especially useful in the case of a very resource consuming decryption procedure, while the combine algorithm is non-time consuming. We also show how to extend the presented MCFE protocol to obtain an OMCFE scheme for the same functionality class.