We introduced a two-layer network model for the study of the immunization dynamics in epidemics. Spreading of an epidemic is modeled as an excitatory process in a Watts-Strogatz small-world network (infection layer) while immunization by prevention of the disease as a dynamic process in a BarabAsi-Albert scale-free network (prevention layer). It is shown that prevention indeed turns periodic rages of an epidemic into small fluctuations, and in a certain situation, actually plays an adverse role and helps the disease survive. We argue that the presence of two different characteristic time scales contributes to the immunization dynamics observed. (c) 2005 Elsevier B.V. All rights reserved.