Permeability is a physical property of rock masses and it is very important in difference field of engineering works, particularly in the Earth science. It is usually calculated with analytical and numerical methods. Analytical methods apply for simple and regular joint set systems and for complicated Discrete Fracture Network (DFN) models only numerical methods are applicable. Numerical simulations demand a large computation resources and great space for data storage and in general it is very time consuming particularly when we have a large model size with densely populated of fractures. Using Effective Medium Theory (EMT) approach for calculation the conductivity and permeability of a lattice network is also reported in the literature as alternative method for numerical solution. However the fluid flow analysis for three-dimensional fracture network using EMT method has not yet been calculated and reported and is the main objective of this research work. Firstly, we developed a code based on the modified EMT methodology for calculating the overall permeability of fracture networks using Matlab program and Representative Elementary Volume (REV) sizes are also approximated. three different aperture patterns are considered; constant aperture, distributed fracture apertures are correlated/uncorrelated with trace length, were set for quantifying the degree of uncertainties between calculated permeability by numerical and EMT methods. Using Monte Carlo simulation techniques, stochastic DFN models were generated with a large number of realizations. The 3Dimensional Distinct Element Code (3DEC) is also facilitated for numerical simulation analysis. The results show that the calculated mean values of permeability between DFN realization models by EMT and numerical methods are same value and only when aperture is constant, result the mean value permeability from EMT method is larger than the mean value permeability from numeric method. The REV size for DFN models with constant fracture aperture is approximated in the same value, however, for uncorrelated aperture- trace length and correlated aperture- trace length models the REV size in EMT method is approximated smaller than numerical methods.