Open leo-barnes opened 2 years ago
There are many features in HEIF that can have privacy implications. In many cases these implications apply to all image formats, but some are specific to the features enabled by HEIF. Many of the features are genuinely useful (storing depth-maps as auxiliary images for instance), so we don't want to forbid them. But it might make sense to have a section on recommended best practices when sharing/editing images.
For example, when sharing or exporting an image for sharing purposes, the following should ideally be done unless the user explicitly does not want it to happen:
- Location data should be stripped
- Other types of identifying metadata should be stripped (i.e. author, unless that is used for copyright)
- Auxiliary images that are not used when rendering the main image (i.e. everything but alpha) should be stripped
- Image overlays should be flattened
- Unless user wants to share all images in a collection, all but the selected image in the collection should be stripped
- If the file contains both a still and video tracks, strip the video unless the user wants to share it
- Thumbnails that do not match the main image should be stripped
- Fully transparent pixels should be zeroed
- Grid padding should be zeroed or replicate the edges
- The contents outside of 'clap' should be zeroed or replicate the edges
There may be other use-cases that could benefit from having recommendations, but sharing seems like the most obvious one.
There are many features in HEIF that can have privacy implications. In many cases these implications apply to all image formats, but some are specific to the features enabled by HEIF. Many of the features are genuinely useful (storing depth-maps as auxiliary images for instance), so we don't want to forbid them. But it might make sense to have a section on recommended best practices when sharing/editing images.
For example, when sharing or exporting an image for sharing purposes, the following should ideally be done unless the user explicitly does not want it to happen:
There may be other use-cases that could benefit from having recommendations, but sharing seems like the most obvious one.