Self-protection is an attribute of autonomic computing systems, reflecting the requirement to proactively defend against attackers, and automatically detect and recover from attacks. As demonstrated by increasing numbers of Internet worms, a single previously unknown vulnerability can cause an entire infrastructure to crumble, due to software and hardware monocultures. One defence against complete failures is diversity: by utilizing differing implementations of software and hardware, the potential total damage from a single exploit is lessened. The self-deployment and selfconfiguration features of an autonomic computing infrastructure make it practical to use diversity as a selfprotection mechanism. We explore the idea of using diversity as a factor in resource allocation decisions, showing how it could be used to limit the damage an attacker can inflict.