We study the complexity of securely evaluating an arithmetic circuit over a finite field ð½ in the setting of secure two-party computation with semi-honest adversaries. In all existing protocols, the number of arithmetic operations per multiplication gate grows either linearly with log |ð½| or polylogarithmically with the security parameter. We present the first protocol that only makes a constant (amortized) number of field operations per gate. The protocol uses the underlying field ð½ as a black box, and its security is based on arithmetic analogues of well-studied cryptographic assumptions. Our protocol is particularly appealing in the special case of securely evaluating a “vector-OLE” function of the form ax+b, where x ∈ ð½ is the input of one party and a, b ∈ ð½w are the inputs of the other party. In this case, which is motivated by natural applications, our protocol can achieve an asymptotic rate of 1/3 (i.e., the communication is dominated by sending roughly 3w elements of ð½). Our implementation of this protocol suggests that it outperforms competing approaches even for relatively small fields ð½ and over fast networks. Our technical approach employs two new ingredients that may be of independent interest. First, we present a general way to combine any linear code that has a fast encoder and a cryptographic (“LPN-style”) pseudorandomness property with another linear code that supports fast encoding and erasure-decoding, obtaining a code that inherits both the pseudorandomness feature of the former code and the efficiency features of the latter code. Second, we employ local arithmetic pseudo-random generators, proposing arithmetic generalizations of boolean candidates that resist all known attacks.
CITATION STYLE
Applebaum, B., Damgård, I., Ishai, Y., Nielsen, M., & Zichron, L. (2017). Secure arithmetic computation with constant computational overhead. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10401 LNCS, pp. 223–254). Springer Verlag. https://doi.org/10.1007/978-3-319-63688-7_8
Mendeley helps you to discover research relevant for your work.