well... I believe that the dot pattern clearly brings rough 3D shape information mainly to ensure there is a real face put in front of the camera (and not a simple fake printed picture on paper). However, the plain NIR image of the face must be the most interesting one to analyze for Face ID for authentification purpose since NIR light does penetrate 0.3 - 1.0 mm within skin = NIR face image must be most interesting one for features identification purpose. It is more or less confirmed by this video where a printed face on textile t-shirt (or face printed on plastic mask) is used to attempt to trick the Face ID system: however, NIR image of printed face on textile or plastic mask cannot reflect typical skin reflectance in NIR.... since it will rather reveal the raw plastic material base itself, the NIR light penetrating through the printed layer which will be transparent if not thick enough in NIR rnage... Therefore it is for sure not the right approach as is to trick the face ID method. One should study NIR face images as a starting point to understand what the Face ID camera is using as baseline for features identification. Then print this well understood NIR face image on a plastic mask with a material as close as possible to skin reflectance (e.g. some wax or silicon type with well selected pigments?) and as homogeneous and thick (1mm) as possible to ensure the NIR light that penetrate the first layer material's doesn't reach the plastic raw material base of the mask itself. Lot of fun for iFixit and others to explore all these scenarios... and lot of interesting future mask printing services for the ones that will succeed to trick this effectively :)
well... I believe that the dot pattern clearly brings rough 3D shape information mainly to ensure there is a real face put in front of the camera (and not a simple fake printed picture on paper). However, the plain NIR image of the face must be the most interesting one to analyze for Face ID for authentification purpose since NIR light does penetrate 0.3 - 1.0 mm within skin = NIR face image must be most interesting one for features identification purpose. It is more or less confirmed by this video where a printed face on textile t-shirt (or face printed on plastic mask) is used to attempt to trick the Face ID system: however, NIR image of printed face on textile or plastic mask cannot reflect typical skin reflectance in NIR.... since it will rather reveal the raw plastic material base itself, the NIR light penetrating through the printed layer which will be transparent if not thick enough in NIR rnage... Therefore it is for sure not the right approach as is to trick the face ID method. One should study NIR face images as a starting point to understand what the Face ID camera is using as baseline for features identification. Then print this well understood NIR face image on a plastic mask with a material as close as possible to skin reflectance (e.g. some wax or silicon type with well selected pigments?) and as homogeneous and thick (1mm) as possible to ensure the NIR light that penetrate the first layer material's doesn't reach the plastic raw material base of the mask itself. Lot of fun for iFixit and others to explore all these scenarios... and lot of interesting future mask printing services for the ones that will succeed to trick this effectively :)
ReplyDelete