Avatar for squadette

#facebook is embedding tracking data inside photos you download. I noticed a structural abnormality when looking at a hex dump of an image file from an unknown origin only to discover it contained what I now understand is an IPTC special instruction. Shocking level of tracking..” https://twitter.com/oasace/status/1149181539000864769



 ‎· You and Mie-Grüneisen 6

вот когда разным людям, которые фотку скачали, будут разное внедрять, будет совсем весело.

 ‎· восторговед 8

@teplorod: а еще не?

 ‎· MI5 ELI5 3

@teplorod: когда людям будут внедрять, будет всяко весело

 ‎· misha

Говорят, что уже давно: "on 2014-06-12, Facebook started inserting a tracking identifier in the metadata" [https://www.hackerfactor.com/blog/index.php?/archives/726-Facebook-Tracking.html].

 ‎· Taivo Lints

В условном 2000 было принято думать, что стеганографически вписывать что-то в невинные картинки будут шпионы и террористы. А тут даже стеганографии нет — видимо, дело будущего (или мы просто не замечаем).

 ‎· 9000 2

FB current use case: Someone uploads an image that isn't tagged. It then gets automatically tagged by FB. Other people save the image.. The tag remains. Someone reports original image as offensive and a human moderator flags it for removal. Based on tag of offensive image, any other images with same tag are treated as being the same offensive image, and are all automatically removed without human intervention, and without anyone reporting them as offensive. This means, if you are part of a private FB group that shares offensive images, and all members of the group have made an agreement not to report any members of the group, your images can still be easily caught by FB and your group members can end up in FB jail for re-uploading them, and your group can be shut down if it has a disproportionately high number of flagged tags in its image shares....and it's all automated, after a human moderator sets a tagged image as offensive.

 ‎· April ᛃ 4

^ That sounds like a silly approach though. I mean, if they only care about exact dupes, why not just match on the md5 of the original? If that metadata persists across edits, that would mean blocking the image even after a user cropped out the offending parts? And in any case, it's pretty easy these days to index the offending content and remove the near duplicates without resorting to watermarks, thus dramatically increasing recall.

 ‎· rychev 1

@app103: I’m afraid this scenario requires giving FB some benefit of doubt which they seem to have defaulted on.

 ‎· псы в рапиде 9

1 2 3 4 5 6 7 8 9 10