ABSTRACT
In this chapter, we address the big-picture issues for knowledge discovery in adversarial settings. We first consider what sort of traces or signals adversaries are likely to leave in data. Several aspects are different from conventional knowledge discovery: adversaries are rare to begin with; they are actively trying to hide and, worse still, to manipulate the process; there is often a need, or at least a desire, for prevention rather than prosecution; and there is almost never a clearly defined, agreed-upon, or even knowable form in which the traces might come.
