Using Machine Learning to Increase NPC Fidelity with Dynamic Preferences Used in Forward Looking Decisions
Abstract
This report summarizes our experience in modeling decision-making preferences within a distributed system of simulated users for cybersecurity training and exercise1. It also provides detail of our explorations of machine learning solutions and how we ultimately enabled increasingly life-like computer activity to autonomous agents over time. In addition, the particular problem we discuss is well-suited to others looking to build an introductory implementation of machine learning, in that our solution requires neither significant investment of money, resources, or data. We hope that operations teams will consider the problems they encounter in the same spirit as our approach, whereby the solution to the problem continues to improve itself in an automated fashion over time.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 18, 2021
- Accession Number
- AD1126910
Entities
People
- Dustin D. Updyke
- Geoffrey B. Dobson
- Thomas G. Podnar
Organizations
- Carnegie Mellon University