This thesis aims to perform a reliability analysis of the stability of 2D soil slopes when they are submitted to water infiltration due to the rains.The time variation of the soil matric suctions is calculated first. The Finite Element Method is used to transform the Richards differential equation into a system of nonlinear first order equations. The nonlinearity of the problem is due to the use of the characteristic curve proposed by van Genuchten (1980). The Modified Picard Method is applied to solve de time-dependent nonlinear equation system. The responses of the flux-problem are transferred to the stability problem in some instants using the same time-interval (normally days).To estimate the stability of the slopes, limit analysis is used. The limit analyses are performed based on the Inferior Limit Theorem of the Plasticity Theory. The problem is defined as an optimization problem where the load factor is maximized. The equilibrium equations are obtained via Finite Element discretization and the strength criterion of Mohr-Couomb is written in the conic quadratic space. Therefore, a SOCP (Second Order Conic Programming) problem is generated. The problem is solved using an interior point algorithm of the code Mosek.Since the soil properties are random variables a reliability analysis can be performed at each instant of the time-dependent problem. In order to perform the reliability analyses, Response Surfaces for the failure function of the slope are generated. In this work, the Stochastic Collocation Method is used to generate Response Surfaces. The Simulation Monte Carlo Method and the FORM (First Order Reliability Method) are used to obtain both the reliability index and the probability of failure of the slopes.Reliability analyses of the Coos Bay Slope in the state of Oregon in USA and in the Vista Chinesa Slope in Rio de Janeiro, Brazil, are presented because they collapse due to rainfall infiltration. The results show that the soil slope fails when the related reliability index is close to two.