Abstract: The perception of the self is a theme much addressed in visual and emotional semiotic literature, psychology (think Gestalt current) and art. Self-portraits are expressive forms that emphasize usually indexical features, creating a continuum - albeit mediated - with the objects from which they spring: the human figure and usually the face. Let us take a leap: new technologies centered on facial recognition are developing programmes to limit recognition, or rather, facilitate anonymity, especially for privacy reasons. Camouflage is a proposal from the Polytechnic, at the root of a collaboration with Unito, which uses Latent Diffusion Models for Attribute-Preserving Image Anonymisation: it is trained on public data sets, as it includes Stable Diffusion as it core and creates anonymised images. After studying and comparing it with FACETS colleagues, I propose to follow with EUFACETS and combine the two points mentioned above (self-perception and technology, specifically Camouflage) and subject the model to feedback from people (part of elder generation) whose portraits have been anonymized. If the re-identification rate is thought to be around 2% (i.e. in 2% of cases the identity of the person can be traced from the image shown), we ask ourselves: does this rate hold if we limit the search to self-perception? Do only 2 out of 10 people re-identify? What elements and levels are involved to prove or disprove identity preservation? The general probabilistic question of the Camouflage method will be questioned and proven by means of specific and targeted self-perception case studies, part of EUFACETS.