Biometric systems are nowadays employed across a broad range of applications.\nThey provide high security and efficiency and, in many cases, are user\nfriendly. Despite these and other advantages, biometric systems in general and\nAutomatic speaker verification (ASV) systems in particular can be vulnerable to\nattack presentations. The most recent ASVSpoof 2019 competition showed that\nmost forms of attacks can be detected reliably with ensemble classifier-based\npresentation attack detection (PAD) approaches. These, though, depend\nfundamentally upon the complementarity of systems in the ensemble. With the\nmotivation to increase the generalisability of PAD solutions, this paper\nreports our exploration of texture descriptors applied to the analysis of\nspeech spectrogram images. In particular, we propose a common fisher vector\nfeature space based on a generative model. Experimental results show the\nsoundness of our approach: at most, 16 in 100 bona fide presentations are\nrejected whereas only one in 100 attack presentations are accepted.\n