Accurate Traffic congestion, beyond its economic and mobility impacts, poses a significant and systemic security risk to urban transportation networks by reducing their resilience to disruptions and amplifying the consequences of incidents. While predictive models excel at forecasting traffic states, they fall short of diagnosing the underlying risk mechanisms, leaving security vulnerabilities unaddressed. To bridge this gap, this paper proposes a security-oriented Four-Factor Congestion Risk Framework that conceptualizes dynamic risk through the lenses of Hazard (probability and intensity of congestion), Exposure (system usage level), Vulnerability (susceptibility to disruptions), and Mitigation Capacity (adaptive and recovery capability). We instantiate this framework into HiST-Graph, a risk-aware Spatio-Temporal Graph Neural Network. Unlike conventional models, HiST-Graph dynamically learns latent risk propagation pathways and disentangles the contributions of the four security-related factors. Extensive experiments on real-world datasets demonstrate that HiST-Graph not only achieves superior predictive accuracy but, more critically, provides interpretable insights into congestion genesis and evolution. The model identifies high-risk segments, quantifies systemic vulnerabilities, and reveals precursor signals to congestion breakdowns. This work offers a paradigm shift from describing congestion to diagnosing its root causes, with direct implications for enhancing transportation security through proactive risk assessment, targeted vulnerability reduction, and informed mitigation capacity planning.
Li et al. (Mon,) studied this question.