Birthday Hunting

The Birthday Problem has a number of applications to incident response. Existing tools can both narrow the focus of the incident response team and limit their experience to a small subset of alerts. This leaves specialized tools to do the analysis before anything is investigated, imposing a range of biases. We show the use of randomly selected investigation of nodes in the environment has a significant likelihood of finding the adversary. This allows for the evaluation of threat hunting and security operations. The approach is then extended to the evaluation of cybersecurity machine learning products. These products may be complicated and opaque. The approach presented avoids the need to understand the internals, shifting analyst focus to business as usual operations.