3 ms·
> The take from this is that they can potentially track photos outside of their own platform with a disturbing level of precision about who originally uploaded
by shawnz 7y ago
> The take from this is that they can potentially track photos outside of their own platform with a disturbing level of precision about who originally uploaded the photo (and much more).
How is it any more powerful in that regard than just hashing the photo?
- shakna 7y agoUploaded photos on many sites undergo postprocessing that may change the hash of the image, different compression technique, etc. without modifying the EXIF data.
- shawnz 7y agoNot if you use a perceptual hash, which is a well established technique that Facebook already uses for other purposes
- doubleunplussed 7y agoFurthermore, plenty of platforms will remove EXIF data. I made a LaTeX equation rendering script that embedded the source LaTeX in EXIF data, the idea being that you could edit the equation after the fact without having to save the source. But practically every tool I used would strip the EXIF data - so I moved to storing the source in the least significant few bits of the pixel values. This worked much better with the places I was uploading the images to (google docs mostly). Obviously fragile to image format conversion though.
- zild3d 7y agoLove that idea for LaTeX. For anyone interested, storing information/messages inside another file is called steganography (https://en.wikipedia.org/wiki/Steganography https://en.wikipedia.org/wiki/Steganography)
- mceachen 7y agoUnfortunately, most phashes are trivial to circumvent. Imperceptibly rotating saturation or leveling brightness will result in a new phash. (If you've found a robust, patent-unencumbered phash, please share, I'd love to use it for PhotoStructure!)