Radio astronomical observations are increasingly corrupted by radio frequency interferences (RFIs), and real time filtering algorithms are becoming essential. In this article, it is shown how spatial processing techniques can limit the impact of the incoming RFIs for phased array radio telescopes. The proposed approaches are based on estimation of the RFI spatial signature. It requires the diagonalization of either the classic correlation matrix or the cyclic correlation matrix of the array. Different diagonalization techniques are compared. Then, RFI detection and RFI filtering techniques are illustrated through simulations on data acquired with the Low Frequency Array Radio telescope, LOFAR. The originality of the study is the use of the cyclostationarity property, in order to improve the spatial separation between cosmic sources and RFIs.