This paper presents a novel framework that leverages the synergistic potential of Graph Neural Networks (GNNs) and Reinforcement Learning (RL) to enhance the cybersecurity of critical infrastructure networks. By modeling network topologies as graphs and applying GNNs to extract embeddings, we enable the detection of vulnerabilities and emerging threats. The integration of RL allows for adaptive and dynamic strategy optimization, ensuring that defense mechanisms can evolve in response to the ever-changing landscape of cyber threats. The proposed framework is validated through extensive simulations and demonstrates significant improvements in the resilience and security of critical infrastructures.