Coverage Directed test Generation (CDG) is a technique that aims to automate the generation of simulation stimuli based on coverage information. This paper presents a novel approach to CDG which is based on inductive learning from examples, in particular Inductive Logic Programming (ILP). The ILP-based CDG methodology takes tests and associated coverage data, some relational background knowledge and a coverage task as learning goal. It produces rules which describe the general structure of tests that achieve the target coverage task. As a first step a rediscovery experiment has been conducted with the aim to show that, given a set of pseudo-randomly generated tests together with their coverage, it is possible to induce rules, at least one per coverage task reached, that correctly characterize the features of tests to target the achieved coverage. The success of this rediscovery experiment confirms the validity of the ILP-based CDG methodology. It establishes the foundations for automatically closing functional coverage metrics using ILP.