Abstarct: In Iran, most of the coal reserves are located deep in the ground or they are very thin in laxyer. To extract and use these sources of coal, Underground Coal Gasification (UCG) seems to be the best option. One of these coal reserves is coal-bearing strata of Mazino area in Tabas. Among the factors and indices that affect the feasibility and economy of the UGC is gas leakage through joints and fractures developed around an underground geo-reactor, i.e. its isolation. Given the significance of this issue, the present study focuses on gas leakage modelling baxsed on the 3D Discrete Fracture Network (DFN) model, since it has not been studied so far. Accordingly, using field data of joints in Mazino area, a 3D geometrical-stochastical modeling of a rock mass jointed network is presented by DFN-FRAC3D code. As regards, this program is developed to consider various distribution functions of joint aperture as well as the relationship between joint length and aperture. Also, the program is developed for detection of connected joint networks. In the following, the connected joint networks are converted to equivalent pipe networks. Finally to study the fluid flow within this network, Numerical software of flow modeling, WaterGEMS that has the capability to make complex pipes and receive other hydraulic information, is used. In order to test the performance of the provided program, a 3D hydraulic model for fracture network of rock mass in Mazino region is presented. It can be concluded that in the case of the execution of UCG in the Mazino coal mine, the resulting gases will not leak to the ground. For the purpose of sensitivity analysis, it is assumed that the seismicity intensity of the region is increased by 50%. In this case, since the number of conductive joints and consequently the number of pipes is very high, the decision was made to apply hydraulic REV size to the area. Accordingly, hydraulic DFN model and then equivalent pipe network model for cubic block dimensions with 3, 5, 8, 10 and 15 m edges were prepared. The pipe network models prepared using WaterGEMS commercial software are analyzed and the flow intensity values are calculated. In this case, the REV size is estimated to be 8 m × 8 m × 8 m. The results show that the average mass flow rate of the studied rock mass is 212 (lit /s). Finally, the results are sensitive analyzed to changes in reservoir pressure, Temperature and Joints physical properties.