Bayesian Network Modeling of Cross-Phase Driver Readiness in Automated Vehicle Cyber-Attacks

Automated vehicles shift drivers into supervisory roles, where engagement in non-driving-related tasks (NDRTs) may degrade driver readiness. This vulnerability becomes critical in adversarial contexts, where cyber-attacks often manipulate in-vehicle information rather than vehicle dynamics. This study examined whether pre-attack driver readiness conditions affect subsequent cyber-attack response performance. Forty-five participants completed a two-phase automated driving experiment consisting of an NDRT phase followed by six cyber-attack scenarios. Situation awareness (SA), visual attention, physiological arousal, workload, and response time were measured. A static Bayesian Network modeled joint probabilistic dependencies across contextual, driver readiness, and behavioral variables. NDRT demand was linked to pre-attack readiness, post-attack workload, and response time in parallel; pre-attack SA was linked to workload but not to response time.