Beyond Trust Management: Counterfactual Mission Provenance Intelligence for Explainable UAV Swarm Security

Autonomous unmanned aerial vehicle (UAV) swarms execute mission-critical tasks through distributed sensing, communication, routing, and control. Existing trust-management approaches can assign a numerical trust score to a UAV, yet they rarely explain why trust changed, which evidence chain caused the change, or how an individual UAV contributed to mission degradation. This paper introduces ProvTrust-UAV, a mission-provenance trust-management framework for explainable UAV swarm security and causal accountability. The framework formalizes a typed Mission Provenance Graph (MPG), computes bounded adaptive trust from behavior, communication, provenance integrity, and mission contribution, and estimates Mission Impact Attribution (MIA) through counterfactual interventions and approximate Shapley-style contribution. To make explainability measurable rather than decorative, the paper defines fidelity, stability, faithfulness, compactness, and operator interpretability metrics for trust-chain explanations. The proposed framework incorporates a structural causal model, a Monte Carlo convergence analysis for Shapley approximation, a probabilistic false-trust reduction analysis, and a sensitivity analysis of trust-weight parameters. Controlled synthetic mission-event simulations over 120 scenarios and 2400 event windows indicate that ProvTrust-UAV improves macro-F1, root-cause attribution accuracy, and false-trust reduction compared with Bayesian, fuzzy, blockchain, deep-learning, and graph-trust baselines. The paper explicitly treats these results as first-stage computational validation and provides an anonymized additional package with simulation summaries, a public-dataset feature-mapping template, and a reproducible scaffold for external validation.