Combinatorial optimization is a fundamental problem found in many fields. In many real life situations, the constraints and the objective function forming the optimization problem are naturally distributed amongst different sites in some fashion. A typical approach for solving such problem is to collect all of this information together and centrally solve the problem. However, this requires all parties to completely share their information, which may lead to serious privacy issues. Thus, it is desirable to propose a privacy preserving technique that can securely solve specific combinatorial optimization problems. A further complicating factor is that combinatorial optimization problems are typically NP-hard, requiring approximation algorithms or heuristics to provide a practical solution. In this paper, we focus on a very well-known hard problem -- the distributed graph coloring problem, which has been utilized to model many real world problems in scheduling and resource allocation. We propose efficient protocols to securely solve such fundamental problem. We analyze the security of our approach and experimentally demonstrate the effectiveness of our approach.