Control-Flow Integrity (CFI) means that the execution of a program dynamically follows only certain paths, in accordance with a static policy. CFI can prevent attacks that, by exploiting buffer overflows and other vulnerabilities, attempt to control program behavior. This paper develops the basic theory that underlies two practical techniques for CFI enforcement, with precise formulations of hypotheses and guarantees.