It's a reverse proxy that presents a PoC challenge to every new visitor. It shifts the initial cost of accessing your server's resources back at the client. Assuming your uplink can handle 300k clients requesting a single 70kb web page, it should solve most of your problems.
For science, can you estimate your peak QPS?
Crazy how I remember the HN post where anubis's blog post was first made. Though, I always thought it was a bit funny with anime and it was made by frustration of (I think AWS? AI scrapers who won't follow general rules and it was constantly giving requests to his git server and it actually made his git server down I guess??) I didn't expect it to blow up to ... UN.
It was frustration at AWS' Alexa team and their abuse of the commons. Amusingly if they had replied to my email before I wrote my shitpost of an implementation this all could have turned out vastly differently.
Also didn't expect you to respond to my comment xD
I went through the slow realization of while reading this comment that you are the creator of anubis and I had such a smile when I realized that you commented to me.
Also, this project is really nice, but I actually want to ask, I haven't read the docs of anubis but could it be that the proof of work isn't wasted / it can be used for something (I know I might get downvoted because I am going to mention cryptocurrency, but nano currency has a proof of work required for each transaction, so if anubis actually does the proof of work as by nano standards, then theoretically that proof of work could atleast be some useful)
Looking forward to your comment!
The only way I see anything like that incorporated is a folding@home kind of thing that could help humanity as a whole.
Of course, if someone makes it work like you suggested, and it catches on, I will personally haunt your dreams forever. Don't give them any ideas.
It's a pretty effective attack because you get large numbers of individual browsers to contribute. Hosters don't care, so unless the site owners are technical enough, they can stay online quite a bit.
If they work with Referrer Policy, they should be able to mask themselves fairly well - the ones I saw back then did not.
Digging it up: https://www.washingtonpost.com/news/the-switch/wp/2015/04/10...
Very similar indeed. The attacks I witnessed where easy to block once you identified the patterns (referrer was visible and they used predictable ?_=... query parameters to try and bypass caches), but very effective otherwise.
I suppose in the event of a hot war, the Internet will be cut quickly to defend against things like the "Great Cannon".
The first involved requests from 300,000 unique IPs in a span of a few hours. I analyzed them and found that ~250,000 were from Brazil. I'm used to using ASNs to block network ranges sending this kind of traffic, but in this case they were spread thinly over 6,000+ ASNs! I ended up blocking all of Brazil (sorry).
A few days later this same web server was on fire again. I performed the same analysis on IPs and found a similar number of unique addresses, but spread across Turkey, Russia, Argentina, Algeria and many more countries. What is going on?! Eventually I think I found a pattern to identify the requests, in that they were using ancient Chrome user agents. Chrome 40, 50, 60 and up to 90, all released 5 to 15 years ago. Then, just before I could implement a block based on these user agents, the traffic stopped.
In both cases the traffic from datacenter networks was limited because I already rate limit a few dozen of the larger ones.
Sysadmin life...