A binary differential evolution algorithm with greedy packaging(GPBDE) is proposed to solve high-dimensional knapsack problems(KPs). A repair strategy with greedy packaging is developed to handle infeasible individuals. To improve population diversity and the ability of global search, a dual transformation is applied to a small number of individuals with lower fitness. To verify the optimization ability of GPBDE, four high-dimensional KPs are used in numerical experiments, and four peer algorithms are also compared with GPBDE. Experimental results show that the GPBDE is superior to other algorithms in finding excellent solutions and handling constraints, and exhibits a fast convergence speed.