Many signal processing applications such as detection, estimation or classification require a statistical model, for the signal of interest, the noise or the interference. Traditionally, the Gaussian modeling is used for its well fitting to the data and for its simplicity. However, nowadays, new challenges in the different systems appear, such as the increase of resolution, the huge amount of data to process… These involve new problems in the processing since the Gaussian model can be not valid anymore. These problems encompass the mismodeling, missing data or outliers present in the data. To answer these relevant questions, robust approaches have been proposed and are more and more used in the signal processing community. The aim of this special session is to provide advanced robust techniques to solve such problems.