Permeability laboratory measurements under in-situ pressures, temperature and strain have been performed on three different diabase alteration facies (metadiabase, chloritized diabase, epidosite) from the Troodos Ophiolite, Cyprus. This aims to study the relations between hydrodynamics, deformation and hydrothermal reaction in the sheeted dyke complex beneath oceanic ridges. The use of water as pore fluid in these experiments favors hydrothermal fluid-rock interactions. All experiments, performed with a Paterson gas-medium apparatus, were achieved at 400 °C, 100 MPa of confining pressure and 50 MPa of pore fluid pressure, conditions at the base of sheeted dykes. Permeability was measured by injection of water and argon before, during and after coaxial deformation. Resulting textures and mineralogy were studied by microscopy and X-ray microtomography in order to identify mineral reactions and to calculate the permeability by numerical simulation after decompression. During stress loading, a compaction/dilatant evolution is observed only in experiments on epidosite. Failure tends to increase permeability by one order of magnitude. For example, using water as pore fluid, permeability measurements after macroscopic failure give the following variations from 4 × 10- 20 m2 to 2.9 × 10- 19 m2 for metadiabase, 1 × 10- 20 m2 to 2.6 × 10- 19 m2 for chloritized diabase and 6.5 × 10- 19 m2 to 3 × 10- 18 m2 for epidosite. Textures suggestive of self-healing and sealing explain permeability reduction by hydrothermal reaction after macroscopic failure. Paradoxically, even using argon as pore fluid, hydrothermal reaction is possible in metadiabases due to dehydration of chlorite. Moreover, fractures appear much finer in rocks enriched in chlorite because of formation of gouge structures. After experiments, 3D images of fracture networks enable to calculate permeability by numerical simulations which values are 4-6 orders of magnitude higher than the experimental measured values. Such results demonstrate that the geometry and the textures observed in unloaded samples are not suitable to estimate permeability and have to be very carefully interpreted.