Pareto optimality of cooperative linear-quadratic difference games in multiobjective dynamic optimization is investigated using a policy iteration based data-driven approach. Different from the existing literature on the multiobjective optimization that heavily depends on the exact model of the dynamic system, the cooperative linear-quadratic difference game with the completely unknown system parameters is well characterized. Firstly, a novel model-free reinforcement learning iterative algorithm is proposed to solve the N algebraic Riccati equations corresponding to the cooperative difference game. Secondly, the convergence of such a model-free algorithm is rigorously guaranteed if the data of the state and the control input is collected enough satisfying the rank criterion. Moreover, Pareto optimal strategies and Pareto optimal solutions are derived via the weighting method combined with an off-policy iterative algorithm. Finally, an off-policy iteration algorithm for the multiobjective cooperative linear-quadratic difference game is presented, where the effectiveness is verified by a numerical example.