Device to device (D2D) communications are emerging as an essential part of technological solutions to boost data rates in the next generation networks. Cognitive radio (CR) opportunistically utilises spectrum to boost spectral efficiency. CR-assisted D2D networks will bring the benefits of both D2D as well as CR together in futuristic cellular networks. This study proposes to opportunistically use TV spectrum white spaces. A joint user selection, mode assignment, and power allocation in CR-assisted D2D networks can definitely yield higher data rates. The proposed study maximises data rate together with users' selection fulfilling various users' power, base station's transmit power, quality of service, and interference related thresholds. This problem is mixed integer non-linear programming and considered non-deterministic polynomial time (NP)-complete. Due to the discrete variables in the problem, finding an optimal solution with the help of an exhaustive search algorithm (ESA) becomes very challenging. The problem gets exponentially complex with the increasing number of user pairs. Thus, the need of another method becomes imperative that yields near optimal solution. Mesh adaptive direct search (MADS) algorithm is considered for solution in the CR-assisted D2D network resource management problem. Simulation results using MADS yield near optimal solution confirming the suitability of MADS for CR-assisted D2D networks.