From Biological to Computational Arms Races – Studying Cyber Security Dynamics

Abstract

The design of computational arms races can draw upon the compelling inspiration of biological arms races. To study cyber security attack-defense dynamics, we have abstracted a description of biological adversarial ecosystems to design an adversarial computational system. The system has elements and processes with abstracted biological analogs. It centers on engagements. Engagements feature adversarial actors (predator/attacker, prey/defender) competing with conflicting objectives culminating in a measurable performance-based outcome. Adversarial dynamics are controlled by coevolution, which selects for better adversaries over multiple engagements using aggregate engagement performance as fitness. Altogether, this system abstracted from nature is capable of population-based, arms race dynamics arising from interacting, evolving adversaries.