SpoofTrackBench: Interpretable AI for Spoof-Aware UAV Tracking and Benchmarking

TL;DR

Radar-based tracking for UAVs is vulnerable to adversarial spoofing, risking mission safety and integrity. SpoofTrackBench offers a modular, reproducible benchmarking framework that injects drift, ghost, and mirror spoofing into the Hampton Skyler Radar dataset and evaluates JPDA and GNN trackers with interpretable overlays. The framework separates clean and spoofed streams, computes drift-from-truth and assignment metrics, and enables cross-tracker comparisons under identical spoof scenarios. This work enables rigorous, open benchmarking of spoof-aware tracking pipelines and sets the stage for AI-driven spoof classification and hybrid tracking architectures.

Abstract

SpoofTrackBench is a reproducible, modular benchmark for evaluating adversarial robustness in real-time localization and tracking (RTLS) systems under radar spoofing. Leveraging the Hampton University Skyler Radar Sensor dataset, we simulate drift, ghost, and mirror-type spoofing attacks and evaluate tracker performance using both Joint Probabilistic Data Association (JPDA) and Global Nearest Neighbor (GNN) architectures. Our framework separates clean and spoofed detection streams, visualizes spoof-induced trajectory divergence, and quantifies assignment errors via direct drift-from-truth metrics. Clustering overlays, injection-aware timelines, and scenario-adaptive visualizations enable interpretability across spoof types and configurations. Evaluation figures and logs are auto-exported for reproducible comparison. SpoofTrackBench sets a new standard for open, ethical benchmarking of spoof-aware tracking pipelines, enabling rigorous cross-architecture analysis and community validation.

Figures (7)

• Figure 1: Schematic comparison of spoof-aware trajectory tracking. JPDA and GNN respond differently to adversarial spoofing within gating regions: JPDA dilutes spoofed detections probabilistically, while GNN applies threshold-based rejection. The illustration highlights trajectory drift, spoof infiltration, and gating logic under adversarial conditions.
• Figure 2: Comparative schematic of JPDA and GNN tracking under spoofing conditions. JPDA handles spoofed detections via probabilistic dilution within the gating region, while GNN applies threshold-based rejection to isolate spoofed inputs. The illustration highlights differences in trajectory drift response and spoof filtering mechanisms.
• Figure 3: Trajectory tracking methods: (a) The GNN trajectory tracking, and (b) The JPDA trajectory tracking.
• Figure 4: Evaluation of trackers: (a) Evaluation of GNN trajectory tracker, and (b) Evaluation of JPDA trajectory tracker.
• Figure 5: Evaluation of trackers under Ghost Spoofing: (a) Evaluation of GNN trajectory tracker, and (b) Evaluation of JPDA trajectory tracker.