Abstract- Dynamic Security Assessment (DSA) is essential for ensuring the reliability and stability of modern power systems, especially under increasingly complex operating conditions driven by renewable integration, variable demand, and frequent topological changes. Traditional simulation-based DSA methods, while accurate, are computationally intensive and unsuitable for real-time applications. Recent advances in machine learning (ML) have shown promise in accelerating DSA, but most existing models overlook the structural dynamics of the power grid, limiting their accuracy and generalization. This paper proposes a novel topology-aware machine learning framework for DSA that explicitly incorporates the power system’s network topology into the learning process using graph-based representations. By leveraging Graph Neural Networks (GNNs) and dynamic topology encoding, the framework captures the spatial and relational dependencies among grid components, enabling robust performance under varying operating scenarios and network configurations. The model is trained and validated on standard IEEE test systems under diverse fault and contingency conditions. Results show significant improvements in accuracy, adaptability, and computational efficiency compared to traditional ML-based approaches. The proposed framework offers a scalable and intelligent solution for real-time security assessment, making it highly relevant for next-generation power system operations.