An inference-proof approach to privacy-preserving horizontally partitioned linear programs

Mangasarian (Optim. Lett., 6(3), 431–436, 2012) proposed a constraints transformation based approach to securely solving the horizontally partitioned linear programs among multiple entities—every entity holds its own private equality constraints. More recently, Li et al. (Optim. Lett., doi:10.1007/s11590-011-0403-2, 2012) extended the transformation approach to horizontally partitioned linear programs with inequality constraints. However, such transformation approach is not sufficiently secure – occasionally, the privately owned constraints are still under high risk of inference. In this paper, we present an inference–proof algorithm to enhance the security for privacy-preserving horizontally partitioned linear program with arbitrary number of equality and inequality constraints. Our approach reveals significantly less information than the prior work and resolves the potential inference attack.