Adversarial design means building a system on the assumption that an opponent will actively probe its weaknesses. Rather than treating users or conditions as always cooperative, designers consider how an attacker might exploit interfaces, bypass safeguards, or adapt when defenses change.
In cybersecurity, this approach informs threat modeling, secure software development, and penetration testing. Teams identify likely attack paths, then test controls and revise the design before weaknesses are exploited in the wild. Defenses may also aim to make attacks harder or less rewarding, not merely absorb them. As with experimental armor in combat robotics, an unusual protective idea is only a hypothesis until it is tested against realistic attacks; effectiveness must be demonstrated, not assumed.