Learning Personalized Human Decision Models in Cyber Defense

Summary

Researchers propose a kernel-based inverse learning framework that infers user-specific latent cognitive traits, such as risk tolerance, curiosity, and prospect, from limited behavioral data, modeling human decisions as a Markov decision process. They validated it with a psychological survey of 108 participants in the TTCP CAGE Challenge 2 testbed, finding it accurately recovers individualized traits and predicts behavior well even with sparse data and behavioral variability.