This paper presents a methodology for general non-linear reliability problems. It is based on dividing the failure domain into a number of appropriately selected subregions and calculating the failure probability as a sum of the probabilities for each subregion. The probability of each subregion is calculated as a product of factors, which can be estimated accurately by a relatively small number of samples generated according to the conditional distribution corresponding to the particular subregion. These samples are generated through Markov Chain Monte Carlo simulations using a slice-sampling-based algorithm proposed by the authors. The proposed method is robust and is suitable for high-dimensional problems. This is in contrast to popular importance sampling methods that often break down for high-dimensional problems. The method is found to be significantly more efficient than Monte Carlo simulations. The efficiency of the method is demonstrated with two examples involving 4000 and 1501 random variables.