Modeling Adaptive Cybersecurity Behavior

Phishing resilience in smart cities depends on how users adapt over repeated attacks, yet most models still treat decisions as static. This study presents PMT-X, a dynamic agent-based extension of Protection Motivation Theory (PMT) in which agents update self-efficacy and cyber awareness while observing peer performance. Using 500 agents per model over 50 decision steps and a dataset of 10,000 phishing and legitimate URLs, the author compared PMT-X- with a static PMT baseline. In the post-training window (steps 21 to 50), PMT-X agents recorded a mean of 11.38 clicks versus 12.78 under PMT, an 11% reduction in risky clicks (incidence rate ratio = 0.89, 95% confidence interval [0.85, 0.93]), despite slightly higher baseline click rates (9.31 versus 9.08). Sensitivity analysis identified response efficacy and self-efficacy as the strongest behavioral drivers. PMT-X provides a transparent test bed for exploring adaptive phishing behavior and tuning targeted cyber awareness interventions.