Can face morphs be detected using face recognition systems?

When the first face morphs were published, it was shown that some commercial face recognition systems generated comparison scores for a morphing attack that were comparable to genuine face comparisons. This means that morphing attacks cannot be distinguished from genuine comparisons. Our recent research shows that other face recognition systems behave differently and generate scores for morphing attacks that lie between genuine and imposter comparisons. In this presentation we explore various ways to make use of these properties in order to make standard face recognition methods more resistant against morphing attacks.