It is shown that coupling nuclear magnetic resonance (NMR) 1D-imaging with the measure of NMR relaxation times and self-diffusion coefficients can be a very powerful approach to investigate fluid infiltration into porous media. Such an experimental design was used to study the very slow seeping of pure water into hydrophobic materials. We consider here three model samples of nuclear waste conditioning matrices which consist in a dispersion of NaNO3 (highly soluble) and/or BaSO4 (poorly soluble) salt grains embedded in a bitumen matrix. Beyond studying the moisture progression according to the sample depth, we analyze the water NMR relaxation times and self-diffusion coefficients along its 1D-concentration profile to obtain spatially resolved information on the solution properties and on the porous structure at different scales. It is also shown that, when the relaxation or self-diffusion properties are multimodal, the 1D-profile of each water population is recovered. Three main levels of information were disclosed along the depth-profiles. They concern (i) the water uptake kinetics, (ii) the salinity and the molecular dynamics of the infiltrated solutions and (iii) the microstructure of the water-filled porosities: open networks coexisting with closed pores. All these findings were fully validated and enriched by NMR cryoporometry experiments and by performing environmental scanning electronic microscopy observations. Surprisingly, results clearly show that insoluble salts enhance the water progression and thereby increase the capability of the material to uptake water.