This research focuses on the optimization of river dikes using probabilistic methods. Its aim is to develop a generic method that automatically estimates the failure probabilities of many river dike cross-sections and gives the one with the least cost, taking into account the boundary conditions and the requirements that are set by the user. Even though there are many ways that may provoke the dike failure, the literature study showed that the failure mechanisms that contribute most to the failure of the typical Dutch river dikes are overflowing, piping and inner slope stability. Based on these, the most important design variables of the dike cross-section dimensions are set and following probabilistic design methods, the probability of failure of many different dike cross-sections is estimated taking into account the abovementioned failure mechanisms. Different cross-section configurations may all comply with a set target probability of failure. Of these, the cross-section that results in the lowest cost is considered the optimal. This approach is applied to several representative dikes, each of which gives a different optimal design, depending on the local boundary conditions. The method shows that the use of probabilistic optimization gives more cost-efficient designs than the traditional partial safety factor designs.