We show the existence of indistinguishability obfuscators (iO) for general circuits assuming subexponential security of: (a) the Learning with Errors (LWE) assumption (with subexponential modulus-to-noise ratio); (b) a circular security conjecture regarding the Gentry-Sahai-Waters' (GSW) encryption scheme and a Packed version of Regev's encryption scheme. The circular security conjecture states that a notion of leakage-resilient security, that we prove is satisfied by GSW assuming LWE, is retained in the presence of an encrypted key-cycle involving GSW and Packed Regev.

Machine learning on encrypted data has received a lot of attention thanks to recent breakthroughs in homomorphic encryption and secure multi-party computation. It allows outsourcing computation to untrusted servers without sacrificing privacy of sensitive data. We propose a practical framework to perform partially encrypted and privacy-preserving predictions which combines adversarial training and functional encryption. We first present a new functional encryption scheme to efficiently compute quadratic functions so that the data owner controls what can be computed but is not involved in the calculation: it provides a decryption key which allows one to learn a specific function evaluation of some encrypted data. We then show how to use it in machine learning to partially encrypt neural networks with quadratic activation functions at evaluation time and we provide a thorough analysis of the information leaks based on indistinguishability of data items of the same label. Last, since several encryption schemes cannot deal with the last thresholding operation used for classification, we propose a training method to prevent selected sensitive features from leaking which adversarially optimizes the network against an adversary trying to identify these features. This is of great interest for several existing works using partially encrypted machine learning as it comes with almost no cost on the model's accuracy and significantly improves data privacy.

Attribute-based encryption (ABE) is a cryptographic primitive which supports fine-grained access control on encrypted data, making it an appealing building block for many applications. In this paper, we propose, implement, and evaluate fully automated methods for proving security of ABE in the Generic Bilinear Group Model (Boneh, Boyen, and Goh, 2005, Boyen, 2008), an idealized model which admits simpler and more efficient constructions, and can also be used to find attacks. Our method is applicable to Rational-Fraction Induced ABE, a large class of ABE that contains most of the schemes from the literature, and relies on a Master Theorem, which reduces security in the GGM to a (new) notion of symbolic security, which is amenable to automated verification using constraint-based techniques. We relate our notion of symbolic security for Rational-Fraction Induced ABE to prior notions for Pair Encodings. Finally, we present several applications, including automated proofs for new schemes.