You’re discussing how to ban people, this isn’t the problem.

The problem is this: In the last hour, 10,000 images were uploaded. Some of those contain CSAM. Now, you have 1 hour to find all the CSAM photos (0 to 10,000 of them). In the next hour, another 10,000 images will be uploaded, some of them containing CSAM…

Unless you have a lot of human moderators, you’re going to use automated tools and get false-postives or false-negatives.

A site like 4 chan banning whole regions isn’t a great example of handling this well. I don’t think I need to explain (but maybe I do) that one person in a region who is posting CSAM doesn’t mean the entire region posts CSAM. You could just opt to block all regions by pulling the site off the internet. Not to mention, does this now mean that 4 chan allows CSAM for certain regions? Yikes. “Children can be abused only in these countries” “I’m sorry but your countries laws prevent images of children being abused, so this content is banned”. Yikes.

maybe look into IP and region banning to prevent someone from just making new accounts.

Again, the technical issue isn’t on banning. Here’s the code to ban user at IP 1.2.3.4:

if ( $_SERVER[‘REMOTE_ADDR’] === ‘1.2.3.4’ ) { die(‘nope!’); }

Here’s the code to ban a user at a specific region (pseudocode):

$geoip = new GeoIPDB(); $region = $geoip->get_region( ‘1.2.3.4’ ); if ( $region === ‘USA’ ) { die(‘nope!’); }

This isn’t difficult.

Now, for the code to DETECT CSAM:

look for skin tone tints (take into account all skin tone colors), look for quantity of skin on image (this would make close-ups of arms possible nude detections), detect a person in the photo, determine the person’s age by the photo… don’t detect images of art or of artful nudes, etc.. or you know this is a lot of work, let’s make the humans detect instead.

Things are setup the way they are because it’s the best way that admins (not just of lemmy instances but of major sites like reddit and facebook) have found to handle these situations.

You could take it a step further and give law enforcement their own backdoor to your site, as Facebook has done, but I would not advocate for that solution. We are in a special place in the internet where we can somewhat self-police our own content, assuming we actually self-police our own content. The way we do this is the way these admins are currently handling this.

It may be reasonable to think that sites like reddit and facebook have it all figured out, but all they have is similar code to what lemmy has, but with a bit more money to pay some content moderators on trust and safety to actually remove this content before users get a chance to see it. The difference between those sites and lemmy is $$$ and that’s not something that’s likely to change anytime soon.