Improving and Measuring Learning Effectiveness at Cyber Defense Exercises

摘要

Cyber security exercises are believed to be the most effective training for the training audiences from top professional teams to individual students. However, evidence of learning outcomes is often anecdotal and not validated. This paper focuses on measuring learning outcomes of technical cyber defense exercises (CDXs) with Red and Blue teaming elements. We studied learning at Locked Shields, which is the largest unclassified defensive live-fire CDX in the world. This paper proposes a novel and simple methodology, called the "5-timestamp methodology", aiming at accommodating both effective feedback (including benchmarking) and learning measurement. The methodology focuses on collection of timestamps at specific points during a cyber incident and time interval analysis to assess team performance, and argues that changes in performance over time can be used to evidence learning. The timestamps can either be collected non-intrusively from raw network traces (such as pcaps, logs) or using traditional methods, such as interviews, observations and surveys. Our experience showed that traditional methods, such as self-reporting, fail at high-speed and complex exercises. The suggested method enhances feedback loop, allows identifying learning design flaws, and provides evidence of learning value for CDXs.