SaTC: CORE: Medium: Secure Resource Management in NextG Radio Access Networks: Attacks, Defenses, and Proofs of Service

About This Grant

Next Generation (NextG) networks aim to create an immersive multi-sensory user experience by supporting new technologies such as integrated precision sensing, artificial intelligence/machine learning (AI/ML) optimizations for automated network management, and a virtualizable architecture running on commodity hardware. Prominent in NextG networks is a new service-centric model, in which infrastructure providers virtualize the network into logically isolated network slices, running services with different demands. This flexible architecture is driven by the deployment of powerful ML algorithms that manage resources at fine and longer timescales. However, the security of decision-making algorithms is under-explored. This project aims to fill this research gap by investigating the security and verifiability of resource allocation for NextG networks. The project’s novelties are in exploring new threats emerging from the automated nature of decision-making and devising robust, secure, and verifiable resource management solutions. The project's broader significance and importance are in improving the availability and self-healing capabilities of the nation's wireless infrastructure and the safety-critical applications it supports. Moreover, the project strengthens the US workforce by providing training opportunities in the critical areas of cybersecurity, AI/ML, and communications. The research agenda is organized in three interrelated thrusts. The first thrust establishes a comprehensive threat model against reinforcement learning-based methods, which are commonly employed for resource allocation, and studies the impact of attacks. Guided by the insights gained from exploring the attack surface in Thrust 1, the second thrust designs robust resource allocation methods for operating in uncertain and maliciously distorted radio environments. A multi-pronged defense, which limits information leakage from the wireless medium while developing attack-resilient allocation methods through adversarial training and robust reward function design, is investigated. The third thrust focuses on verifying that allocation policies adhere to the service-level agreements between stakeholders by building proofs of service to validate the resource transactions that take place over the air. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.