The quantification and modeling of fluid flow in fractured rocks are extensively studied to solve and predict numerous economic or environmental problems (hydrothermalism, geothermy, storages, etc.). Indeed, discontinuities such as faults and fractures are potential sites for fluid circulation and have important implications for the hydraulic properties of rocks. The matrix permeability of igneous rocks is generally small and, consequently, the global permeability is mostly controlled by the fault and fracture networks. Therefore, the quantification of the fractured rock hydraulic properties strongly depends on the knowledge of the geometrical parameters of fractures (orientation, extension, aperture, density) and of the final 3D modeling of the fracture network organization. In the specific case of the Soultz-Sous-Forêts geothermal reservoir, a new statistical analysis of the fault and fracture networks is proposed to precise the actual 3D structural model of the reservoir (Sausse et al., 2009; Dezayes et al., 2009). The statistical characterization of the fractures and faults is realized with the re-interpretation of the whole U.B.I. images database available at Soultz. 1800 fractures are determined along the three deep Soultz well paths, grouped into main conjugates fractures sets, showing a mean N-S orientation and a mean dip of 70°, consistent with the Oligocene N-S extension responsible of the formation of the French Rhine graben. A correlation between the geometric parameters of fractures, width W and extension L is proposed and follows a power-law type correlation of the form L = k . WD with k, a coefficient characteristic of the facies and D the fractal dimension of the fracture set. These parameters are used to determine the volumetric density of fractures (number of fractures/m3) at the wells scale. Finally, this density and the statistics of fracture properties are used to constrain stochastic simulation of a discrete fracture network (DFN) in the geothermal reservoir.