I don't believe there are any logs being transferred, just the abusive IPs are shared with the central DB.

So if an IP starts hitting a ton of rules (like .env access, repeated 403s, 404s, 429s, etc.. Or specific AppSec rules) then that IP is blocked and sent to their central DB where it's pushed out to everyone running crowdsec.

On my setup crowdsec has been more effective than cloudflare at stopping scanners and bots from overloading things, cloudflare just lets basically everything through so I stopped using it.

That said I'm curious if you remember the source of the real world tests, because I'd love to see what's actually better.

You're not wrong. If CrowdSec is your entire approach, it's a little weird. But I use it as a (small) part of my overall blocking strategy. I use other lists and frankly, block most IPs by default. Those other IPs aren't counted in the email screenshot though... only the CrowdSec based ones are coming through. I think a lot of the 137k above are just scanning bots to be frank.

Do you have a source on the real world tests?

Is there an alternative you suggest? I use Crowd-Sec in part because I don't know of anything else that does the same job. That is, blocks malicious IPs without snooping on all traffic content (looking at you, Cloudflare).

Sure. So OpnSense is the real gem here. Or really any solid firewall/edge device CrowdSec is an optional (free) add-on module in OpnSense. It's immediately operational at a minimal level without any registration. However, if you choose to register within the CrowdSec UI then you can load up more lists and tie it to your instance.

I also load up Hagezi's Threat Intelligence Blocklists into my OpnSense firewall rules, which get updated (by him) on a rolling 12 hour basis. I also have OpnSense query Github every 12 hours to pull the updates.

It is pretty good I use it on many machines. You install it on your machine with a reverse proxy, it then analyses the traffic logs for malicious activity or for known bad IPs and then creates iptable block rules. The web interface is the cloud interface that it connects to, because as the name implies, it works by using log data from all members to block bad actors efficiently