Advancing End-to-End Programmable Networks: Exploring the Interplay of Generative AI With Opportunities and Challenges

The rapid proliferation of Internet of Things (IoT) devices and the continuous evolution of network communications are reshaping the global network landscape, demanding advanced management techniques to address the growing complexity of data and service requirements. Generative Artificial Intelligence (GAI) has emerged as a transformative solution, enabling multi-modal data analytics and decision-making. However, traditional network architectures lack the scalability and flexibility required to accommodate the diverse resources and compute-intensive demands of GAI. To address these challenges, we propose a novel end-to-end programmable network framework that provides scalable, flexible support for enhanced storage, computation, and transmission capabilities. This architecture empowers compute-intensive and data-intensive applications by leveraging specialized GAI models designed for network optimization, dynamically enhancing resource allocation and operational efficiency. Furthermore, we implement a GAI-assisted framework to improve DDoS detection within programmable networks. By using diffusion models for synthetic data generation, the framework addresses data scarcity in privacy-constrained environments and simplifies GAI model deployment through knowledge distillation. Simulation results validate the proposed framework’s effectiveness, demonstrating significant improvements in detection latency and detection accuracy.