If I have several backends that more or less depend on each other anyway (for example: Lemmy + pict-rs), then I will create separate databases for them within a single postgres - reason being, if something bad happens to the database for one of them, then it affects the other one as well anyway, so there isn’t much to gain from isolating the databases.

Conversely, for completely unrelated services, I will always set up separate postgres instances, for full isolation.

That particular instance was very recently the source of a lot of CSAM and spam, so that’d be why. A lot of instances recently upped their security to combat that.

Just to add some more context, there was an attacker recently who created accounts on several Lemmy instances and used those accounts to spread CSAM. On lemm.ee, this attacker created 4 accounts over a 24h period, but was not able to upload any CSAM to our servers due to our stricter upload rules (we require 4 week old accounts to upload any images at all), and all of the 4 accounts were removed very shortly after creation (most of them within an hour of signing up). The attacker gave up trying to use lemm.ee very quickly, and moved on to other instances.

I just wanted to share this context to illustrate that while indeed the different measures we implement to protect the instance can have a negative impact on legitimate users, I really believe that overall, they have a net positive effect. In addition to Cloudflare DDoS protection and image upload restrictions, we also have a separate content-based alerting layer on top of Lemmy, which allows our admins to quickly notice when something suspicious is going on. As another example, this alerting has allowed us to extremely efficiently deal with a current ongoing spam attack on the Fediverse, and I bet many lemm.ee users aren’t even aware of this attack due to the quick content removal. We will continue to improve our defenses, and hopefully try to limit the impact on real users as much as possible, but some trade-offs are necessary here in order to protect the overall userbase.

The nice thing about Lemmy is that you can always host your own instance, even if it’s only for your own individual use. You can basically use your own instance as a proxy - other instances will not see how or from where you are connecting to your instance.

Large instances are being attacked almost constantly at this point in smaller and bigger ways. Almost all measures we implement to combat these attacks come with some trade-offs for the rest of the userbase.

Important note, this feature is only available for US customers.

Nowadays it’s allowed only for users with >4 week old accounts. It’s not perfect, but having this barrier to entry will hopefully prevent at least some problems.

I think the OP is talking about Lemmy having both a content preview and a text area for link posts.

Some users tend to write their own summary in the text area, so when opening up a post, the result will be:

• Large title written by the OP
• Automatic preview of the page, generated by Lemmy
• Text summary of the page, written by the OP

I agree that this is a bit clunky in terms of UX

What exactly is the issue with our admins? If you feel you’ve received some unjustified moderation, feel free to contact me and I can have a look.

As a test, I ran this on a very early backup of lemm.ee images from when we had very little federation and very little uploads, and unfortunately it is finding a whole bunch of false positives. Just some examples it flagged as CSAM:

• Calvin and Hobbes comic
• The default Lemmy logo
• Some random user’s avatar, which is just a digital drawing of a person’s face
• a Pikachu image

Do you think the parameters of the script should be tuned? I’m happy to test it further on my backup, as I am reasonably certain that it doesn’t contain any actual CSAM

Any thoughts about using this as a middleware between nginx and Lemmy for all image uploads?

Edit: I guess that wouldn’t work for external images - unless it also ran for all outgoing requests from pict-rs… I think the easiest way to integrate this with pict-rs would be through some upstream changes that would allow pict-rs itself to call this code on every image.