I use an iptables rule like:

    # connection limit for HTTP
    -A INPUT -i eth0 -p tcp --syn --dport www -m connlimit ! --connlimit-above 128 --connlimit-mask 24 -j ACCEPT

This basically means, "only accept requests to port 80 if the /24 has fewer than 128 open connections". It is to kill things like Slowloris, but you can bring the limit down and kill scrapers. You can also block based on connection rate:

    -N RATE_CHECK
    -A INPUT -p tcp -m multiport --dports www -m state --state NEW -j RATE_CHECK
    -A RATE_CHECK -m recent --set --name RATE
    -A RATE_CHECK -m recent --update --seconds 60 --hitcount 4 --name RATE -j REJECT
    -A RATE_CHECK -p tcp -m multiport --dports www -j ACCEPT

This will deny any connection that is the 5th connection or higher in a single minute. I use this rule for ssh and smtp, but some tweaking might be adequate for a web server.

(Note: all these rules are for default deny with a rule like "-A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT" to ignore established states. If your firewall setup is different, then the rules will need to be modified.)

There is also a program called fail2ban that will read logs and apply temporary bans based on what the logs say.

My solution to this problem is to just cache agressively. Hundreds of requests per second is nothing for Varnish.

be careful with things like this, it could get you locked out of your server very quick.

denyhosts and fail2ban are better ideas for dropping bad actors on the net.

It's not a problem; the rules only apply to eth0, but I connect over tun0. And, it's time-limited to 60 seconds either way. And, I have web console access.

So this is not something I worry about. ssh password crackers and script kiddies taking down my website are a little more worrisome.

Do you have a document on your site establishing that this idiot is violating your acceptable use policy? Then you should report them to Amazon's abuse group:

http://aws-portal.amazon.com/gp/aws/html-forms-controller/co...

You can use technical countermeasures to defend your site, but you'd be doing the rest of the Internet a favor to get them knocked off Amazon.

We do. I'll send that over to them. Great idea!

Have you heard back from Amazon? We've got a similar incident going on, which I reported last Friday (3 business days ago), but haven't yet seen any acknowledgment from them.

Feed them rss containing uniquely generated strings. Then google for the unique strings.

Also, block generic user-agents on rss feeds. Most everyone was able to change theirs for delicious, and we did an insane amount of rss traffic.

I found this to work very well.

Only thing is you need to append keys randomly, say one out of 5 requests, in different locations if possible. And you should hand out 20 or so different unique strings.

Then use google alerts to find them republishing your stuff an send DMCA takedown notices with Google (don't send them a c&d first so you have the element of surprise and maximum chance of causing real pain). If they don't take care of the DMCA issue with Google, Google will remove their pages from the index. They won't know why but their site will slowly die.

google alerts = clever.

the other thing to do is include some some encoding of the served-to IP address (maybe md5 of the first and last half of the IP?)

To generate the unique string, do SHA1(content + TheirIP + unique secret). The chance of collision is practically zero. Also, you can then conclusively prove that it was you who created this unique string - which may come in handy.

(Talking about SHA1 because I used it myself in a lighter version - just a hash of my name - to check how the search engines work with the content of my little blog).

I like it, but you need to do this in a way that you append a known/deterministic string so you can monitor THAT one in Google Alerts.

I don't care whether the unique strings collide (although I love the idea of being able to tie the IP address of the crawler to the page on which the content was published for maximum evidence collection), I just want Google Alerts to find them reliably AND I don't want the spammer to catch on and be able to easily remove the unique strings before I catch him.

The only way for him to find my little landmines (to strip them) would be to read through every piece of content he scraped from me.

By the way - thinking of all this - why would not you turn this scraper into a free advertisement for you ? If you can detect when it crawls your site, just insert randomly the backlinks to your site into the content you give out to them.

This way you will want that they steal more from you :-)

(how to put these backlinks in a way that would be difficult to remove - is another story. But if you use other links, e.g. tinyurl, in your material - then you could probably use that :-) Still fairly simple to remove but would require more work from them.)

Not true. See http://en.wikipedia.org/wiki/Birthday_problem

The birthday problem basically just cuts the number of bits in half. For SHA1 this changes from a chosen collision at absurd odds to a random collision at merely ridiculous odds.

Feed them ads with affiliate codes in them.

That's a fun idea. Thanks.

I think this is a good approach for identifying who this is, but you should assume good faith. They probably don't even know they're causing you problems. Try asking them nicely to limit their request rate, and only block them if they're not willing to do that.

Interestingly I don't see any mention of your robots.txt, or if the scraper identified themselves in the UA string. If the person or organization running the scraper is legit, they would probably like to know their software is misbehaving, so that they can fix the issue. Did you attempt to contact them before launching into a cat and mouse game?

Yeah, definitely try contacting them first. Even if their software is being idiotic, they might be a fellow startup or someone who's a bit technically dim but has reasonable intentions. Of course, if you get the brush off, then you have carte blanche for raising hell ;-)

You can use varnish to cache all of these requests and just never expire them. Make the page as small as possible. It'll make the overhead of serving the requests as low as possible.

http://www.varnish-cache.org/

You can also use Varnish ACLs to block the EC2 IP addresses.

And if they have a really obvious user-agent you could block that using Varnish: http://omninoggin.com/web-development/block-unwanted-spam-bo...

And if you're not afraid of using Varnish's VCL InlineC stuff, then you could add rate limiting to Varnish: http://drcarter.info/2010/04/how-fighting-against-scraping-u...

Basically... Varnish has become my Swiss army tool for rejecting crap traffic (and caching good traffic obviously).

Amazon previously kept a semi-official list of all netblocks for EC2 in their forum. They've recently moved to posting updates as official announcements, here is the latest:

https://forums.aws.amazon.com/ann.jspa?annID=877

It's interesting to see the size of their deployments measured in usable IP space.

If something like this can slow your servers down, you're probably going to be in serious trouble when you get an inbound link from a heavily trafficked site. I'd recommend you put a caching proxy like Varnish in front of your web server process.

Varnish serves 404s quickly and happily, and will keep the storm from even reaching your apaches. You won't even notice it when the scrapers come in.

You can simply contact Amazon AWS and report them. Having run yottaa.com live for the last 6 months our monitoring nodes have been reported and if they are causing problems we fix them.

We are try to provide valuable service to people and would want to know that we are causing problems. We have implemented the ability to work with robots.txt files.

If the nodes are not from a legitimate service then they will be shutdown.

The proper way to fix this is to add a per-IP rate limiter to the entire site. Most other fixes are just a cat-n-mouse game.

If you're feeling nefarious, 301 their requests to a static, empty RSS feed.

If you can figure out who's actually running it (does the user agent say?), simply sending them an email and asking them to throttle their script may be the simplest solution.

Agent doesn't say unfortunately.

You should report the behavior to Amazon:

http://aws-portal.amazon.com/gp/aws/html-forms-controller/co...

In my experience they take it seriously and deal with it promptly.

Depends on the technology you're using to serve the rss feeds. I'd consider writing something to tail my access logs and automatically update firewall rules when a host generates a certain number of 404's in a certain period of time.

It shouldn't be too difficult to write a fail2ban rule to do this.

mod_rewrite does exactly what you want here. Rewrite urls matching that pattern to a static file, so the request returns quickly and has very little overhead.

I assume you're running Apache? Have you taken a look at ModSecurity?

I found it a little tricky to set up (the core rules it ships with by default are way too aggressive, IMHO), but it does exactly what you're asking: block requests by regex.

We're on a hybrid nginx/apache setup.

I just added some lines to my nginx.conf though that appear to be working based on tips from here and elsewhere:

location ~* /neighborhoods/[-\w]+/feed/?$ { deny all; }

this is why I love HN, all the suggestions have been actually constructive and nobody is suggesting 4channy nonsense