Among mountainous permafrost landforms, rock glaciers are mostly abundant in periglacial areas, as tongue-shapedheterogeneous bodies. Passive seismic monitoring systems have the potential to provide continuous recordings sensitive to hydro-mechanical parameters of the subsurface. Two active rock glaciers located in the Alps (Gugla, Switzerland and Laurichard, France) have then been instrumented with seismic networks. Here, we analyse the spectral content of ambient noise, in order to study the modal sensitivity of rock glaciers, which is directly linked to elastic properties of the system. For both sites, we succeed in tracking and monitoring resonance frequencies of specific vibrating modes of the rock glaciers during several years. These frequencies show a seasonal pattern characterized by higher frequencies at the end of winters, and lower frequencies in hot periods. We interpret these variations as the effect of the seasonal freeze-thawing cycle on elastic properties of the medium. To assess this assumption, we model both rock glaciers in summer, using seismic velocities constrained by active seismic acquisitions, while bedrock depth is constrained by Ground Penetrating Radar surveys. The variations of elastic properties occurring in winter due to freezing were taken into account thanks to a three-phase Biot-Gassmann poroelastic model, where the rock glacier is considered as a mixture of a solid porous matrix and pores filled by water or ice. Assuming rock glaciers as vibrating structures, we numerically compute the modal response of such mechanical models by a finite- element method. The resulting modeled resonance frequencies fit well the measured ones along seasons, reinforcing the validity of our poroelastic approach. This seismic monitoring allows then a better understanding of location, intensity and timing of freeze-thawing cycles affecting rock glaciers.