Using Homomorphic Encryption for Large Scale Statistical Analysis

The development of fully homomorphic encryption schemes in recent years has generated considerable interest in the field of secure computing. In this paper, we consider the problem of performing statistical analysis on encrypted data. Specifically, we focus on two tasks: computing the mean and variance of univariate and multivariate data as well as performing linear regression on a multidimensional, encrypted corpus. Due to the high overhead of homomorphic computation, previous implementations of similar methods have been restricted to small datasets (on the order of a few hundred to a thousand elements) or data with low dimension (generally 1-4). In this paper, we first construct a working implementation of the scale-invariant leveled homomorphic encryption system in [Bra12]. Then, by taking advantage of batched computation as well as a message encoding technique based on the Chinese Remainder Theorem, we show that it becomes not only possible, but computationally feasible, to perform statistical analysis on encrypted datasets with over four million elements and dimension as high as 24. By using these methods along with some additional optimizations, we demonstrate the viability of using leveled homomorphic encryption for large scale statistical analysis.

[1]  Craig Gentry,et al.  Fully Homomorphic Encryption with Polylog Overhead , 2012, EUROCRYPT.

[2]  Pascal Paillier,et al.  Public-Key Cryptosystems Based on Composite Degree Residuosity Classes , 1999, EUROCRYPT.

[3]  Taher ElGamal,et al.  A public key cyryptosystem and signature scheme based on discrete logarithms , 1985 .

[4]  Chris Peikert,et al.  On Ideal Lattices and Learning with Errors over Rings , 2010, JACM.

[5]  Frederik Vercauteren,et al.  Fully homomorphic SIMD operations , 2012, Designs, Codes and Cryptography.

[6]  Craig Gentry,et al.  A fully homomorphic encryption scheme , 2009 .

[7]  Vinod Vaikuntanathan,et al.  Can homomorphic encryption be practical? , 2011, CCSW '11.

[8]  Michael Naehrig,et al.  ML Confidential: Machine Learning on Encrypted Data , 2012, ICISC.

[9]  Vinod Vaikuntanathan,et al.  Efficient Fully Homomorphic Encryption from (Standard) LWE , 2011, 2011 IEEE 52nd Annual Symposium on Foundations of Computer Science.

[10]  Ronald L. Rivest,et al.  ON DATA BANKS AND PRIVACY HOMOMORPHISMS , 1978 .

[11]  Dan Boneh,et al.  Evaluating 2-DNF Formulas on Ciphertexts , 2005, TCC.

[12]  S. Fienberg,et al.  Secure multiple linear regression based on homomorphic encryption , 2011 .

[13]  Craig Gentry,et al.  Fully Homomorphic Encryption without Bootstrapping , 2011, IACR Cryptol. ePrint Arch..

[14]  Ivan Damgård,et al.  Multiparty Computation from Somewhat Homomorphic Encryption , 2012, IACR Cryptol. ePrint Arch..

[15]  Zvika Brakerski,et al.  Fully Homomorphic Encryption without Modulus Switching from Classical GapSVP , 2012, CRYPTO.

[16]  Craig Gentry,et al.  Homomorphic Evaluation of the AES Circuit , 2012, IACR Cryptol. ePrint Arch..