Fair resource sharing and early-exit AI improve secure edge computing
Proportional-Fair Resource Allocation and Dual-Threshold Early-Exit Inference for Secure Cooperative Multi-Layer Edge Intelligence
Networking and Internet ArchitectureComputer Vision and Pattern RecognitionEmerging TechnologiesMachine Learning
Summary
Handling AI tasks on devices and nearby servers can be tricky when resources and security matter. The authors created a system called FREDI that lets devices decide quickly if they can classify information themselves or need help from a local server. They designed a fair way to share wireless and computing resources while maintaining privacy when important events occur. Their approach balances device workload, communication limits, and accuracy to work well even as many devices connect. The system was tested with popular AI models showing it works fast and fair for many users.
What this means in practice
- •For edge system engineers: Configure edge devices and servers to share wireless and computing resources efficiently for AI inference with privacy safeguards using FREDI's proportional-fair framework.
- •For wireless network planners: Design resource management policies that support secure, cooperative inference tasks with early-exit confidence thresholds on wireless edge networks.
Authors
Thai T. Vu, John Le, Tu N. Nguyen, Jun Shen, Quang Vinh Duong, Ha Nguyen
Abstract
This paper proposes FREDI (Fair Resource Allocation for Edge Dual-Threshold Inference), a secure wireless edge-intelligence framework for event-triggered inference in a cooperative user equipment (UE)--edge server (ES)--cloud system. Each UE performs early-exit convolutional neural network (CNN) screening using dual confidence thresholds, while critical events are securely offloaded to an edge server for detailed classification. We formulate a proportionally-fair utility maximization problem that jointly optimizes UE--ES association, wireless and processing resources, and confidence thresholds. FREDI decomposes the problem into proportional-fair resource allocation and dual-threshold inference optimization. We prove that the detected-critical event set is set-monotone non-increasing in both thresholds, and exploit the finite empirical confidence domain for exact threshold optimization. An empirical resource--utility response envelope yields a computable global suboptimality bound and a sufficient condition for global optimality. By pre-eliminating infeasible UE--ES pairs and exactly projecting out bandwidth and transmit-power variables, the resource-allocation subproblem is reduced to a mixed-integer exponential-cone program solvable to the certified global optimality within a prescribed gap. Numerical results with early-exit MobileNetV2 and ShuffleNetV2 demonstrate near-perfect UE fairness with aggregate utility close to a Sum-Utility benchmark, reveal security-induced resource fragmentation, and demonstrate the Stage-A scalability from 6 to 144 UEs with median solving time below 0.1~s in the tested configurations.