Fracture flow is the dominant flow in many cases in a fractured rock mass, which behaves more heterogeneous, anisotropic and irregular than porous flow because of the complexity of the fracture networks. Rock mass with fracture networks is commonly represented by homogenized continuous models to reduce the impractical computational complexity. In this paper, a density-reduced equivalent discrete fracture network (E-DFN) method is proposed to simplify the original highly dense target DFN (T-DFN) models. An equivalent permeability factor is derived for the fracture sets in the simplified E-DFN models. A number of T-DFNs with different fracture patterns are stochastically generated and fluid flow is simulated in both T-DFN and corresponding E-DFN to test the effectiveness of the proposed method. Results show the permeability similarity of the two DFN systems and high accuracy of the proposed E-DFN method. The proposed method takes the advantage of permeability similarity of DFN systems, and it significantly reduces the computational complexity and cost while the characteristics of the fluid flow such as heterogeneity, irregularity, and discontinuity in highly fractured rock mass are well retained. Because of the statistical equivalence of the two DFN systems, Monte Carlo simulations can be used to optimize the equivalent permeability factor and to achieve more accurate and reliable results.