In this work we propose a kernel-based robust nonparametric approach to direct data-driven control of linear systems, in the presence of bounded noise affecting the measurements data. First we formulate the problem of designing a controller in order to match the behavior of a given reference model. Then, we design the controller by applying previous results by some of the authors in the field of kernel-based nonparametric error-in-variables identification. Finally, we show the effectiveness of the presented technique by means of two simulation examples.