Paper 2016/1071

Functional Encryption from Secure Enclaves

Sergey Gorbunov and Dhinakaran Vinayagamurthy

Abstract

Functional encryption (FE) is an emerging paradigm for public-key cryptography that enables fine-grained access control over encrypted data. In FE, each function (program) $P$ is associated with a secret key $sk_P$. User holding $sk_P$ and a ciphertext $ct$ encrypting a message $msg$, can learn $P(msg)$ in clear, but nothing else about the message is revealed. Unfortunately, all the existing constructions are either very restrictive in the supported classes of functions, or rely on non-standard mathematical assumptions and satisfy weaker security notions such as indistinguishability-based security, or far from satisfying practical efficiency for general function families. In this work, we present a construction of functional encryption in a hardware assisted model of computation. We prove the security of our construction under the simulation-based definition. We present an implementation of our construction and show essential evaluation results, which demonstrate that our construction is very practical. In our evaluation, key-generation, encryption and decryption take around $1$, $22$ and $140$ milliseconds for linear regression programs over 4 million sample points. Our construction is motivated by the recent advances in processors that enable creation of encrypted memory containers.