Chosen Ciphertext Security via Point Obfuscation

In this paper, we show two new constructions of chosen ciphertext secure (CCA secure) public key encryption (PKE) from general assumptions. The key ingredient in our constructions is an obfuscator for point functions with multi-bit output (MBPF obfuscators, for short), that satisfies some (average-case) indistinguishability-based security, which we call AIND security, in the presence of hardto-invert auxiliary input. Specifically, our first construction is based on a chosen plaintext secure PKE scheme and an MBPF obfuscator satisfying the AIND security in the presence of computationally hardto-invert auxiliary input. Our second construction is based on a lossy encryption scheme and an MBPF obfuscator satisfying the AIND security in the presence of statistically hard-to-invert auxiliary input. To clarify the relative strength of AIND security, we show the relations among security notions for MBPF obfuscators, and show that AIND security with computationally (resp. statistically) hard-to-invert auxiliary input is implied by the average-case virtual black-box (resp. virtual grey-box) property with the same type of auxiliary input. Finally, we show that a lossy encryption scheme can be constructed from an obfuscator for point functions (point obfuscator) that satisfies re-randomizability and a weak form of composability in the worst-case virtual grey-box sense. This result, combined with our second generic construction and several previous results on point obfuscators and MBPF obfuscators, yields a CCA secure PKE scheme that is constructed solely from a re-randomizable and composable point obfuscator. We believe that our results make an interesting bridge that connects CCA secure PKE and program obfuscators, two seemingly isolated but important cryptographic primitives in the area of cryptography.