Secure Transformation for Multiparty Linear Programming

With the rapid increase in computing, storage and networking resources, data is not only collected and stored, but also collaboratively analyzed. This creates a serious privacy problem which often inhibits the use of the distributed data. In this paper, we focus on the problem of linear programming, which is the most important sub-class of optimization problems. Specifically, distributed linear programming problems allow different companies (even competitors) to collaboratively seek the maximum profit or minimum cost by better utilizing the combination of their limited resources (constraints). However, the constraints and the objective are typically distributed across different companies. Since the constraints generally refer to internal limitations or capacities, and the objective specifies internal costs or value, serious privacy and security problems arise if these are completely disclosed to other companies. The main contribution of this paper is to introduce privacy-preserving distributed linear programming techniques to resolve the above issue for several real-world distributed linear programming problems. We present a transformation based approach that works for arbitrarily partitioned data distributed between multiple parties, and experimentally evaluate our solution.