Media forensics has attracted a lot of attention in the last years in part\ndue to the increasing concerns around DeepFakes. Since the initial DeepFake\ndatabases from the 1st generation such as UADFV and FaceForensics++ up to the\nlatest databases of the 2nd generation such as Celeb-DF and DFDC, many visual\nimprovements have been carried out, making fake videos almost indistinguishable\nto the human eye. This study provides an exhaustive analysis of both 1st and\n2nd DeepFake generations in terms of facial regions and fake detection\nperformance. Two different methods are considered in our experimental\nframework: i) the traditional one followed in the literature and based on\nselecting the entire face as input to the fake detection system, and ii) a\nnovel approach based on the selection of specific facial regions as input to\nthe fake detection system.\n Among all the findings resulting from our experiments, we highlight the poor\nfake detection results achieved even by the strongest state-of-the-art fake\ndetectors in the latest DeepFake databases of the 2nd generation, with Equal\nError Rate results ranging from 15% to 30%. These results remark the necessity\nof further research to develop more sophisticated fake detectors.\n