Receptors on-board satellites are being implemented to track civil aircrafts all around the world. This new scenario requires novel methods to process the signals in order to efficiently retrieve more updated and reliable position and status data of every aircraft. To reach the required performance, it is indeed needed to engage carefully chosen algorithms of data analysis and processing. Machine learning algorithms, in particular k-nearest neighbors and support vector machines, are employed to estimate the potential success in decodifying ADS-B messages in highly congested areas, and simulations are performed to obtain the training and testing signals. First, the ADS-B communication system is described; second, multivariate analysis and machine learning algorithms are studied. Finally, the results obtained from machine learning methods are compared and future studies are proposed.