Abstract. Self-optimizing agents in many systems achieve competitive advantage by discovering how to exploit their environment and other agents, often maliciously subverting the operating rules of the system. For example, viruses and cancer exploit the reproductive mechanisms of host cells to replicate and spread; bullies use intimidation to exert power, violating social norms; and cybercriminals exploit software vulnerabilities, eroding trust in networked systems. As robotics and massively embedded computation become increasingly pervasive and autonomous, we can expect that evolution and technology will collide in the physical world. Autonomous vehicles and fleets of drones will revolutionize transportation and warfare, computers will be embedded in our home appliances, and medical implants will restore function to paralyzed limbs and deliver drugs. These systems will be guided by adaptive algorithms encoded in software, ultimately becoming more autonomous, more lifelike, and more likely to cheat.
Biological defense systems are ubiquitous, robust, multi-level and exquisitely complex. Similarly, social animals have evolved numerous forms of restraint to prevent conflicts from escalating unnecessarily. We aim to understand how these could arise and be sustained in cyber physical systems. We will examine the evolution of defense and restraint in nature and apply these strategies to software, robots and cyberwarfare.