Workshop on Foundations of Computer Security - FCS'03
Ottawa, Canada, 26-27 June 2003

Oblivious Comparator and its Application to Secure Auction Protocol

Hiroaki Kikuchi (Tokai University - Japan)


Abstract

This paper presents a protocol for Secure Function Evaluation (SFC) in which n players have secret inputs E[a1], E[a2], ... , E[an], of a known boolean function f, and they collaborate to compute the ciphertext of the output of the boolean function, E[f(a1, ... , an)] . The main result is a completeness theorem (Theorem 3.1) which states that an arbitrary function can be evaluated at the oblivious party without help ofprivate information. The proposed protocol is based on the Jakobsson and Juels's Mix-and-Match scheme in which the truth table of a target function is row-wise randomized (mixing) using a Mix network, and then players perform ``matching'' the designated output ciphertext and the corresponding rows. The biggest difference between the proposed SFC and the Mix-and-Match is that the proposed protocol does not require any involution of key holders to evaluate function, while the Mix-and-Match needs key holders to perform threshold decryptions at every step of evaluation of boolean gates. One disadvantage of the proposed scheme is the Reed-Muller expansion involves an exponential blow-up in the number of input, n, as conventional schemes such as CyptoComputer proposed by Sander, Young, and Yung. This paper presents an efficient construction for a primitive called `oblivious comparator' with n-round complexity between the comparator and n players but the bandwidth spent by one communication is independent from n (linear to the size of values to be compared), and hence it does not suffer the blow-up in n. The oblivious comparator is suitable to implement a secure auction because an auctioneer communicates with bidders once at time, and performs evaluation without help of trusted key holders. In addition, the proposed construction allows arbitrary complicated functions including a search for second highest, a resolution the winner, and a dynamic programming (for combinatorial auction).


Iliano Cervesato