This thesis studies the modeling of credit risk in static credit portfolios. In this context one of the most important issues is to understand the dependence between defaults, that is one of the measures that helps shaping the probability distribution of the total loss of a credit portfolio. In many cases this dependence is unknown, and therefore it is hard to find directly the related risk measures (such as the VaR). For this reason, we will try to find the tightest possible bounds for the VaR of the total loss distribution, in particular when marginal probability of default of each obligor is a Bernoulli with probability p and the multivariate Bernoulli vector is exchangeable.