Computing Reviews

Superimposing encrypted data
Yu K., Yu T. Communications of the ACM34(2):48-54,1991.Type:Article
Date Reviewed: 08/01/91

The authors are concerned with the processing of encrypted data. For example, could we take an encrypted amount in hundreds of dollars and convert it to yen (that is, multiply it by a known constant, such as 12,689), thus obtaining an encrypted value of the converted amount, without ever knowing the plaintext amount? This particular conversion is multiplicative, so it could be done under the RSA algorithm [1]. More generally, we might want to transform the encrypted vectors X and Y linearly to obtain rX+sY (where r and s are scalars) in encrypted form without decrypting. The method is to use the time-reversal transformations for the encryption of X={xj(1)}, so xi(t+1)+xi(t−1)=f[{xj(t)}]modk. The indices in f are shifts of the index i, and the transformation is executed a certain number of times starting with xi(0)=0 and xi(1)=plaintext. This algorithm is essentially the DES method [2]. If the vector is of dimension n, then the subscripts are taken modulo n. The trick is to take f to be linear, such as fxi−3+7xi+2 +63xi+5. This limits the choice of f severely, creating a security problem. One modification is to permit more arbitrary backgrounds xi(0). The authors note that for some choices of f, the encryption resembles the finite difference solution of the wave equation, 2 W &slash; ∂ t 2 = c 2 2 W &slash; ∂ x 2 , so the problem is analogous to the superposition of waves.


1)

Rivest, R. L.; Shamir, A.; and Adleman, L. A method for obtaining digital signatures and public-key cryptosystems. Commun. ACM 21, 2 (Feb. 1978), 120–126. See <CR> 19, 7 (July 1978), Rev. 33,278.


2)

Konheim, A. G. Cryptography, a primer. Wiley, New York, 1981.

Reviewer:  Harvey Cohn Review #: CR115247

Reproduction in whole or in part without permission is prohibited.   Copyright 2024 ComputingReviews.com™
Terms of Use
| Privacy Policy