Abstract
This work presents a novel approach for dealing with failures and attacks on Transparent Optical Packet Switching (TOPS) mesh networks. The approach is composed of two phases, whereas the first one dynamically dimensions the resources in the network, the second one applies an incremental learning algorithm that generates an intelligent policy. At each node, such a policy allows a self-healing behavior when there are failures or attacks in the network. Finally, the performance of this approach is presented as well as future research lines.