We will show how methods based on statistical physics and complex networks approaches may help to predict the appearing of crises such as epidemics. These methods also suggest efficient immunization strategies to coop with such crises. The epidemics could occur in social systems as well as in communication networks such computers or cellphones. The methods are based on the percolation theory approach which is extended to complex networks to include more realistic scenarios, such as the limited time of epidemics or the dynamical nature of links. Questions such as how to identify the most crucial spreaders and giving a limited amount of immunization doses, how to prioritize the recipients will be also discussed.