We know that we can decode user-object id/id_str using the snowflake algorithm to reveal some metadata about the account.
I’ve also found that we can extract a similar looking id from the “profile_image_url_https” key & confirmed that it can be decoded similarly to reveal the date/time the avatar was last updated.
This can be useful in looking at large follower lists to expose possible networks of inauthentic (“fake”) accounts.
I think that the snowflake algo started to be universally applied to user-object id/id_str at about the end of Nov 2012, but it would be useful to get that confirmed please.
When looking at update date/times for avatars, I have found some anomalies though, and I would like advice as to how/why
One particular example is avatars showing an apparent common update date of 2013-09-14 on a range of accounts that appear genuine & that have a wide range of account creation dates.
For these accounts the “profile_image_url_https” all contain an id that looks these:
378800000396842837
378800000418844492
378800000420272988
378800000484903300
378800000470287540
Could someone from the Twitter API/dev support team offer any clarity about how the snowflaking works for avatar urls? Is there a decoded date (or a unix timestamp) after which they are always reliable, for instance?
1 Like
I’ve it at 2013-01-22 for 64bit user ids rolled out (before that i think was just experiments, like https://twitter.com/overflow64 ) twitter-history/changes.csv at master · igorbrigadir/twitter-history · GitHub
I think this may be due to twitter changing how images are compressed or something similar? That would be my best guess.
system
Closed
#3
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.
