Over the years, switches and network routers have been compromised frequently, and a lot of vulnerabilities have occurred in network infrastructure. Secure routing (SR) is one of the challenges that currently exists in computer networks. Software-defined networks (SDN) are designed by assuming that routers or switches are trustworthy. In SDN, untrusted devices have resulted in security issues such as traffic analysis, failure to receive information, packet change and removal, inaccurate routing, and network downtime. Using encryption methods is a possible solution to deal with some of these problems, but it requires additional infrastructure as well as significant overhead at runtime. One of the most trusted routing methods is through replicated devices (switches or routers). Recently we have seen less attention given to the number of replicated devices in SR. In this paper, the problem of SR is converted into a multi-objective optimization problem, considering the reliability of different manufacturers and SR with untrusted devices is performed. To this end, a mathematical model is provided to study the objectives of maximum reliability and cost minimization. The NSGA-II algorithm is applied to determine the optimal number of replicated devices in order to minimize the cost of implementing SR in spite of the presence of untrusted devices in SDN. Our simulation results show that our proposed method compared to the base method (without considering optimization) decreases implementation cost by 27% and increases the reliability from 70% to 93.2%.