Security (attacks against ML systems) — is the mirror image of → Abuse (of ML systems)

The paper's own footnote makes the two areas a matched pair defined by swapping which side of an attack the machine learning system sits on: "security" differs from "abuse" in that the former involves attacks against a legitimate ML system by an adversary (e.g. a criminal tries to fool a face recognition system), while the latter involves attacks by an ML system controlled by an adversary (e.g. a criminal trains a "smart hacker" system to break into a website) (concrete-problems, §"Abuse:", p. 21). Both require an adversary and both require an ML system, but security's ML system is the target while abuse's ML system is the weapon. This is also what separates both most sharply from the paper's own subject: accidents-in-ml requires no adversary at all, only "poor design," placing security and abuse at the opposite end of the adjacent-areas list from where the paper's own topic would sit.