Multi-input Functional Encryption with Unbounded-Message Security

Multi-input functional encryption (MIFE) was introduced by Goldwasser et al. (EUROCRYPT 2014) as a compelling extension of functional encryption. In MIFE, a receiver is able to compute a joint function of multiple, independently encrypted plaintexts. Goldwasser et al. ( EUROCRYPT 2014) show various applications of MIFE to running SQL queries over encrypted databases, computing over encrypted data streams, etc. The previous constructions of MIFE due to Goldwasser et al. (EUROCRYPT 2014) based on indistinguishability obfuscation had a major shortcoming: it could only support encrypting an a priori bounded number of message. Once that bound is exceeded, security is no longer guaranteed to hold. In addition, it could only support selective-security, meaning that the challenge messages and the set of "corrupted" encryption keys had to be declared by the adversary up-front. In this work, we show how to remove these restrictions by relying instead on sub-exponentially secure indistinguishability obfuscation. This is done by carefully adapting an alternative MIFE scheme of Goldwasser et al. that previously overcame these shortcomings (except for selective security wrt. the set of "corrupted" encryption keys) by relying instead on differing-inputs obfuscation, which is now seen as an implausible assumption. Our techniques are rather generic, and we hope they are useful in converting other constructions using differing-inputs obfuscation to ones using sub-exponentially secure indistinguishability obfuscation instead.