Our CSAM scanner ("CCS") is now operating in private beta mode with a small number of servers using the service to detect illegal content.
-
Our CSAM scanner ("CCS") is now operating in private beta mode with a small number of servers using the service to detect illegal content. To date, we have processed over 200,000 media files, with 11 matches for known CSAM, and 1 false positive.
Our automated alert system notified our CCS end users, who have taken the illegal content down. IFTAS retains the pertinent media for one year as required by law, and has reported the observations to NCMEC.
Learn more: https://about.iftas.org/activities/moderation-as-a-service/content-classification-service/
-
@iftas some context for the numbers:
200,000 is a very small number, but it does loosely compare to the SIO study (https://stacks.stanford.edu/file/druid:vb515nd6874/20230724-fediverse-csam-report.pdf) that showed 112 matches in 325,000 media files.
Very roughly, we are at 4 per 100,000, SIO showed 34 per 100,000
However, we are working with a very small number of servers, we anticipate the observation rate increasing as we add servers.
Some trivia: we are seeing 1400 media files/hour or ~33k/day, and roughly 1 in 3 posts has media attached.
-
-
@iftas Do you share the hashes ? it might be useful for many other open source projects.
-
The hash databases need to be independently licensed or acquired by any individual party that wants access. Sharing the hashes publicly would immediately invalidate the hash db.
https://connect.iftas.org/library/legal-regulatory/csam-primer/ has some pointers, there are free services that will hash and match for you once you're approved for access.
