Network protection against natural and human-caused,hazards has become,a topical research theme in engineering and social sciences. This paper focuses on the problem ofallocating,limited retrofit resources over multiple highway ,bridges to improve ,the resilience and robustness of the entire transportation system in question. The main modeling challenges,in network ,retrofit problems ,are to capture ,the interdependencies among individual transportation facilities and to cope ,with the extremely high uncertainty in the decision environment. In this paper, we model the network retrofit problem as a two-stage stochastic programming ,problem ,that optimizes a mean-risk ,objective of the ,system loss. This formulation hedges well against uncertainty, but also imposes computational challenges due to involvement ,of integer ,decision variables and increased dimension ,of the ,problem. Anefficient algorithm is developed, via extending the well-known L-shaped method using generalized benders decomposition, to efficiently handle the binary integer variables in the first stage and the nonlinear recourse in the ,second stage of the ,model ,formulation. The proposed,modeling ,and solution methods ,are general and can be applied ,to other ,network design problems as well.