Malware Mimicry Drift: Exploiting the Fitness-Beats-Truth Theorem for AI-Driven Malware Containment Author: George White

Abstract

AI-driven malware represents a paradigm shift in cyber threats, utilizing generative models to dynamically rewrite code and evade traditional signature-based defenses. Inspired by the Fitness-Beats-Truth (FBT) theorem from evolutionary psychology, which posits that natural selection favors perceptual systems tuned for fitness payoffs over objective reality, this paper proposes a novel defense hypothesis: Malware Mimicry Drift (MMD). We posit that by presenting a deceptive, high-fitness interface to malware-a sophisticated honeypot environment-the defender can exploit the malware's inherent reward-function optimization. This forces the malware to adapt its behavior to the illusory environment, a process we term "mimicry drift," leading to its containment and neutralization. We present a quantitative model, the Malware Domestication Rate (MDR), to predict the effectiveness of this approach. A clear falsification condition is defined, and preliminary simulations suggest the hypothesis offers a viable strategy for proactively managing adaptive cyber threats.