Infectious diseases are a major threat to global health. Spatial patterns revealed by epidemic models governed by reaction-diffusion systems can serve as a potential trend indicator of disease spread; thus, they have received wide attention. To characterize important features of disease spread, there are two important factors that cannot be ignored in the reaction-diffusion systems. One is that a susceptible individual has an ability to recognize the infected ones and keep away from them. The other is that populations are usually organized as networks instead of being continuously distributed in space. Consequently, it is essential to study patterns generated by epidemic models with self- and cross-diffusion on complex networks. Here, with the help of a linear analysis method, we study Turing instability induced by cross-diffusion for a network organized SIR epidemic model and explore Turing patterns on several different networks. Furthermore, the influences of cross-diffusion and network structure on patterns are also investigated.