Towards Formal Verification of Federated Learning Orchestration Protocols on Satellites

2024-10-17

Telecommunications Forum

AI Summary

Ensuring temporal correctness and high-probability termination of federated learning (FL) orchestration protocols in mobile edge environments—particularly low-Earth-orbit (LEO) satellite networks—remains challenging due to dynamic topology and time-varying communication links. Method: This paper proposes the first formal verification framework integrating celestial mechanics modeling with stochastic timed automata (STA), comprising two stages: (i) modeling communication dynamics via orbital mechanics constraints, and (ii) statistical model checking using Uppaal SMC to verify deadlock-freedom, termination, and quantify termination probability. Contribution/Results: Unlike conventional CSP-based approaches ill-suited for dynamic topologies, our framework formally guarantees protocol liveness and quantifies reliability under realistic orbital configurations. Experiments show FL task termination probability ≥99.3% and end-to-end latency deviation within stringent space–ground real-time constraints. This work pioneers the integration of astrodynamics into formal verification of distributed learning protocols, establishing a new paradigm for trustworthy collaborative intelligence in space systems.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Python Testbed for Federated Learning Algorithms (PTB-FLA) is a simple FL framework targeting edge systems that implements both centralized and decentralized FL orchestration protocols, which were formally verified in a previous paper using the process algebra CSP. This approach is appropriate for systems with stationary nodes but cannot be applied to systems with moving nodes. In this paper, we use celestial mechanics to model spacecraft movement, and timed automata (TA) to formalize and verify the centralized FL orchestration protocol, in two phases. In the first phase, we created a conventional TA model to prove traditional properties, namely deadlock freeness and termination. In the second phase, we created a stochastic TA model to prove timing correctness and to estimate termination probability.

Problem

• Federated Learning
• Mobile Devices
• Satellite Communication

Innovation

• Astrophysical Dynamics
• Temporal Automata
• Federated Learning on Mobile Devices