In this work we define multiple relaxations to the definition of correctness in secure obfuscation. While still remaining meaningful, these relaxations provide ways to obfuscate many primitives in a more direct and efficient way. In particular, we first show how to construct a secure obfuscator for the re-encryption primitive from the Decisional Learning with Errors (DLWE) assumption, without going through fully homomorphic encryption. This can be viewed as a meaningful way to trade correctness for efficiency.
Next, we show how our tools can be used to construct secure obfuscators for the functional re-encryption and multi-hop unidirectional re-encryption primitives. In the former case, we improve upon the efficiency of the only previously known construction that satisfies the stronger notion of collusion-resistant obfuscation (due to Chandran et al. – TCC 2012) and obtain a construction with input ciphertexts of constant length. In the latter case, we provide the first known obfuscation-based definition and construction; additionally, our scheme is the first scheme where the size of the ciphertexts does not grow with every hop.