Mistaken self-defense presents a puzzle: in at least some cases agents are intuitively justified in imposing defensive harm on an apparent aggressor, despite (in fact) facing no genuine threat. I argue that these cases motivate a more expansive view of the moral grounds of permissible self-defense, allowing that in addition to liability-based permissions, defenders sometimes enjoy vulnerability-based permissions. In particular, when individuals S behave in ways that conventionally signal that they pose a threat to an agent P, S cannot reasonably demand that P refrain from defensive action, so P is permitted to defend herself. I develop an account of what it is to conventionally signal threateningness, and explore some limitations on what signaling behaviors can do the relevant moral work.