Why Your Firewall Might Be Useless Against DDoS Attack

4198 So you've spent some money on a shiny new firewall, and you have it mounted nicely in the rack right next to your ethernet switch(es) and DSL/cable modem. You're all set, right?

Well, maybe not. A company here on Long Island ran into trouble last week when the website they are hosting on a server in their office went down. Yes, the server is in an office with a single Internet connection, but I'm not going to bother with the discussion about how hosting your site on a single-homed network is a bad idea. We'll save that for another day. Turns out that the site was down primarily because of a distributed denial of service (DDoS) attack against the web server. Why was someone attacking the web server? Who knows, but it was causing serious problems for those folks and something had to be done about it. I'll spare you the suspense and let you know right away that one of our affiliated tech gurus was able to solve the problem for them, but it did point out one item that people tend to forget about when it comes to network attacks.

Often even the best firewall in the world will be useless against a DDoS attack. This is because firewalls are really good at filtering and blocking traffic using all kinds of nifty algorithms, but they can't do that until said traffic gets to them.

Lets say that you have a cable modem that runs at 20 Mbps downstream and 5 Mbps upstream. That means that in the very best case, you can move 20 megabits per second worth of data down that pipe and into your home/office/network/whatever. Now lets say that someone decides that you're a good target to attack because they find the website you're hosting in the office to be distasteful in some way. They decide to launch a modestly scaled attack on you, and suddenly your web server is subject to a 65 Mbps torrent of miscreant packets designed to make it feel really badly. Have you seen the problem yet?

The problem in this case is not that there are nasty packets coming your way. Presumably your trusty firewall will know now to block those (if you have it configured correctly - also another story). The problem is the amount of data you have to deal with. That 65 Mbps stream is WAY more than your 20 Mbps cable modem can handle, so while your firewall is happily dropping packet after packet on the floor for you, your pipe is full and nothing can get to you any longer. Filter until you're blue in the face, because your website, and the rest of your office, is now effectively offline. You might as well unplug your cable modem and call it a day because the bad guys have won.

So what can you do? In some cases the attack is not large enough to fill the incoming pipe. Once you've filtered it and rendered it ineffective, the attackers simply go away and move along to the next dirty deed on their to-do list. Luckily this is what happened to our friends last week. A bit of intelligent firewall configuration and they were back in business. In the case where the attack is large enough to fill your pipe, you have no choice but to enlist the aid of your access provider. Those black-hat packets have to be stopped before they get into your pipe, and only your provider can do that. In some instances your access provider can "null route" the IP address under attack. In more complex instances where you have multiple access providers and are running a dynamic routing protocol, you might be able to make this happen automatically using BGP community strings. Sometimes you'll have to call and work your way through the phone maze to find the people that can help. In other cases, your access provider might simply shrug its shoulders and decide that you're on your own. Sadly, this also happened here last week (shame on you, Optimum Lightpath!).

The point is, network attacks are a fact of life on the Internet, so do yourself a favor and plan to spend 20-30 minutes on the phone with your service provider at some point in the near future finding out how they will help you deflect an attack if you ever find yourself in that position. Document the procedure - who needs to be called or emailed, what information they'll need, and how long it might take for them to take action. Having it all written down will go a long way toward calming things down when you're under the gun. Also be sure you're familiar with the DNS for your domains and how its all controlled. When combating a network attack there is likely going to be some DNS changes to be made, so don't wait until things are out of control to learn how to do that, or who can do it for you.

A little knowledge, and a little advance preparation, can go a long way toward making a bad situation easier to deal with.

In future blog postings we'll talk about how site-crippling network attacks can be avoided/handled by having your sites hosted with a hosting/co-location/datacenter provider.

Founded in 1996, MacConnect is the first and largest Mac-Centric ISP on the planet. Providing world class hosting solutions that are as easy as your Mac, MacConnect is the first choice for any Mac user in need of web, email and application hosting. Find us online at MacConnect.com



Tags: Blogs ď All Things infrastructure ď

Login † or † Register † †

Actually, stateful firewalls don’t do much against even trivial DDoS attacks - their state-tables fill up and good traffic is crowded out by programmatically-generated bad traffic which matches all the firewall rules, is well-formed traffic, but is generated by bots.

DDoS attacks are attacks against capacity and/or state - stateful firewalls shouldn’t be placed in front of servers ever, because there’s no state to inspect (incoming connections are all unsolicited), and they will choke and die much quicker than the naked servers themselves.

Policy should be enforced via stateless ACLs in hardware-based routers/layer-3 switches which can handle mpps; source-based remotely-triggered blackholing (S/RTBH) and intelligent DDoS mitigation systems (IDMSes) are tools which can be utilized to dynamically mitigate DDoS attack traffic.

Good comments!  All true.

The point still stands that a well crafted attack on capacity must be dealt with upstream, either manually or through “automagic” blackholing.  If you’re hosting on a consumer/small-business broadband connection, good luck with that.

In the cited case, the customer’s Sonicwall actually did have the ability to filter the attack once it was properly configured to drop (rather than simply log) packets matching any of the built-in DOS signatures.  It was a nice surprise that the box proved to be so capable.  Had they been running the usual $100 SOHO firewall, all bets would have been off, even in the face of what turned out to be a fairly feeble attempt at a DDoS.

It wasn’t a serious attack, in that case - very large 10gb/sec firewalls from major vendors fall down under DDoS all the time.  Sounds as if this was a very basic, low-rate packet-flooding attack with out-of-policy packets which could easily be dropped. 

Blackholing isn’t automagic, nor should it be - auto-mitigation of any kind is a Bad Idea, as it can be gamed.  Manually-triggered S/RTBH and/or manually-engaged IDMS (which handle antispoofing, layer-7 attacks, et. al.) that don’t carry state are definitely the mitigation tools of choice.

I didn’t mean to imply that auto-blackholing should be completely automatic, only that it can/should be done without having to pick up the phone and talk to someone upstream of you.  I do agree that it should only happen after a live person with a clue decides that it needs to be done.

And you are right, it was a pretty lame attack.  Funny, except that these folks were down for over 30 hours while the first “security guru” they hired stared at firewall logs and fiddled with trying to block traffic from any source IP outside of the ARIN jurisdiction.  Hopefully they didn’t pay him much.  wink

One interesting thing is that at one point they tried using a cheap GoDaddy hosting account as bait to move the attack away from their LAN (the attack followed the hostname).  That worked for a few hours until GoDaddy decided that $7.00/month isn’t enough for them to deal with even trivial DoS attacks and turned it off.

Heh, most ‘security’ types focus only on confidentiality and integrity, don’t think about availability, in my experience.  Defending against DDoS is an opsec thing (vs. infosec, which is policymaking/audit), and many organizations don’t have an opsec team.

Follow Us

Twitter Facebook RSS! http://www.joeryan.com Joe Ryan

Most Popular

iPod




iPhone

iLife

Reviews

Software Updates

Games

Hot Topics

Hosted by MacConnect - Macintosh Web Hosting and Mac Mini Colocation                                                    Contact | Advanced Search|