Slashdot Mirror


EveryDNS Under Botnet DDoS Attack

mellow marsh writes "EveryDNS, sister company to OpenDNS (which runs the PhishTank anti-phishing initiative), has been hit by a massive distributed denial-of-service attack. The attack started sometime Friday afternoon and, from all indications, was targeting Web sites that used free DNS management services provided by EveryDNS. At the height of the DDoS bombardment, EveryDNS was being hit with more than 400mbps of traffic at each of its four locations around the world. From the article: '"We were collateral damage," Ulevitch explained... Because law enforcement is involved, Ulevitch was hesitant to release details of the actual target but there are signs that some of the targets were "nefarious domains" that have since been terminated.'" OpenDNS, which makes use of EveryDNS services, was affected for a time, until they spread their authoritative DNS more broadly. The EveryDNS site is now reporting that the attack is continuing but has been mitigated and is not affecting operations.

21 of 154 comments (clear)

  1. puppy by Feyr · · Score: 5, Funny

    /., like kicking a dead puppy.

  2. Re:COM != NET by SaDan · · Score: 4, Informative

    What parent said. The main site is http://www.everydns.net/ not .com.

    Another quality, editor approved Slashdot story. Great job, guys.

  3. correct URL by barista · · Score: 3, Informative

    How about linking to the correct url?

  4. Heh by davidu · · Score: 5, Informative

    The site is EveryDNS.Net.

    I'll keep it up for Slashdot, let me just move it around a bit. :-)

    -david

    --

    # Hack the planet, it's important.
    1. Re:Heh by Anonymous Coward · · Score: 5, Funny

      You must be new...oh

  5. Questions? by davidu · · Score: 5, Informative

    Since I've been getting a lot of questions from folks about EveryDNS, how we've been stable and around so long, how we dealt with this DDoS and how we manage to cover our costs I am writing a response that will probably be posted here on Slashdot tomorrow or Monday to answer all these questions.

    If you have questions about this or DDoS in general, feel free to ask them here and I'll make sure to cover them in my response. I'll be writing about what we've seen and what I generally do when it comes to soaking up traffic and how we handled this event in particular. (The short answer: find the smartest people you can to help you and then start taking corrective action)

    Thanks!

    David Ulevitch

    --

    # Hack the planet, it's important.
    1. Re:Questions? by TubeSteak · · Score: 3, Insightful
      Because law enforcement is involved, Ulevitch was hesitant to release details of the actual target but there are signs that some of the targets were "nefarious domains" that have since been terminated.
      What does that mean?
      Was this a 'righteous' attack on malicious websites?
      Or just some intramural warfare by one nefarious group upon another?
      --
      [Fuck Beta]
      o0t!
    2. Re:Questions? by davidu · · Score: 4, Interesting

      In short, the latter. Nothing is ever righteous when it comes to DDoS. :-)

      --

      # Hack the planet, it's important.
    3. Re:Questions? by Beryllium+Sphere(tm) · · Score: 4, Interesting

      Bless you for offering to answer questions! That sort of cooperation is indispensable if security is going to improve.

      1. How did you manage the response? The one-smart-person-in-charge-who-stays-awake-the-who le-time approach? The small-team-with-independent-responsibilities model? The review-what-happened-at-shift-change model?

      2. What tactics worked, and even more important, what didn't work?

      3. What sort of agreements should people have in place with their upstream ISP prior to an incident?

      4. How intelligent was the attack traffic? Randomized payload? Does anyone bother spoofing addresses any more?

      5. Was it a guided attack or a fire and forget? In other words, did the scum make any changes to their tactics in real time as you tried corrective action?

      6. What if anything can be done in the first few minutes/hours?

      7. If you had to choose between capacity and filtering, which would you choose?

    4. Re:Questions? by davidu · · Score: 3, Informative

      4x400mbps == 1200mbps at times.

      That's less trivial to filter, especially when your upstream isn't being cooperative. In our case, which you'll read about tomorrow or Monday, we quickly were able to jump onto a network run by some folks with very very high levels of clue; nLayer operated by Richard Steenbergen. Their website is cheesy -- don't let it fool you. They are a seriously run network providing transit across the country to a bunch of other networks. Check routeviews for proof.

      -david

      --

      # Hack the planet, it's important.
  6. Real ripple effects, even from this small event. by ScentCone · · Score: 5, Insightful

    A client (a pretty large retail chain) was using EveryDNS for forward lookups to the mail server's A record. Mail they were sending out started to bounce because receiving mail servers weren't happy when trying to validate the sending box. In once case, a vital piece of mail sent to a state taxing authority couldn't get through on a month-end calendar deadline, causing much grief. Yes, alternate communcations channels are always an option, but it wasn't immediately clear why the two mail servers in question appeared to be hating each other.

    Worse, the state government box's spam filtering appliance blacklisted the retailer's server, and a third party admin had to get involved to free things up. Quite a mess.

    But the real lesson? People who say that a "cyber attack" couldn't really hurt the economy are wrong, wrong, wrong. This stuff can be really disruptive, and this was a pissant little scaled-down example. No major damage, but a lot of thrashing around, untold manhours of lost productivity, and (in the case of the anecdote in question, involving just one retail company), probably some tax fines which will require much tail chasing to get waived once the the story is clearly told, assuming the state government in question is feeling sporting about it.

    --
    Don't disappoint your bird dog. Go to the range.
  7. "nefarious domain" is a loaded and subjective term by plasmacutter · · Score: 5, Insightful

    What is "nefarious"?

    to some.. the pirate bay and allofmp3 are "nefarious domains"..

    to others "www.f**Ktimewarner.com" and "walmartsucks.com" are "nefarious domains"

    and to others "www.wikipedia.org" and "www.aclu.org" are "nefarious domains".

    I have a lot of trouble with the idea that DDOS attacks were being carried out in (apparently successful) attempts to wipe domains off the face of the earth..

    this implies the attackers had no legal standing to take those domains offline.. then they call them "nefarious" after the fact.

    --
    VLC FOR MAC IS DYING! IF YOU DEVELOP, PLEASE SAVE IT!!
  8. Botnet? Cal it what it is! by Chris+Tucker · · Score: 4, Insightful

    Compromised Windows machines network.

    Where are the class action suits against Microsoft for continually producing such flawed software that makes it easy to 0wn a box?

    If it wasn't for 20 some years of MS indifference towards security, there wouldn't be botnets like this, being used for DDOS attacks and forwarding billions of spams a day.

    --
    Guaranteed! This comment 100% Anthrax free!
  9. Open Letter to all Trolls by tomstdenis · · Score: 4, Interesting

    You're pricks.

    Nothing positive or lasting will come out of trolling (and yes: this means you anonymous asshats on /. and in usenet).

    So why not be part of a winning team and stop script kiddie'ing around from your parents basement.

    Sincerely,
    The Rest of the Human Race.

    --
    Someday, I'll have a real sig.
  10. Re:Poor engineering? by Anonymous Coward · · Score: 3, Funny

    The problem is, EasyDNS could only afford an AOL dialup to put their servers up. On top of that, the "server", is really just an old Pentium MMX with 32megs of RAM running bind on top of cygwin on top of Windows 95. Unfortunately, the admin let his 16 year old sister use the machine to browse MySpace (and who knows what else), so let's just say the machine is running other "services" as well.

  11. incompetence effects, not ripple effects by SuperBanana · · Score: 3, Insightful

    In once case, a vital piece of mail sent to a state taxing authority couldn't get through on a month-end calendar deadline, causing much grief.

    Maybe a)it shouldn't be left until the deadline and b)sent via email, if it's so damn important.

    And maybe you not tell clients to use a free DNS hosting service as their sole DNS provider...

  12. They deserve the grief by Pig+Hogger · · Score: 3, Insightful
    In once case, a vital piece of mail sent to a state taxing authority couldn't get through on a month-end calendar deadline, causing much grief.
    That grief is well deserved. E-mail is **NOT** reliable, and delivery is **NOT**, **CAN NOT** and **WILL NOT** be guaranteed. So anyone stupid enough to entrust "vital" communication to e-mail rightly deserves to have his arse whipped real good.

    Myself, a month ago I missed an opportunity to collaborate on a TV miniseries. Why? Because the moron who asked me for my collaboration absolutely trusted e-mail, and it was **THE** message that bounced thanks to a network glitch, and that moron didn't think of calling me on the **PHONE**. Well, if they were stupid enough to trust e-mail like that, they probably would have made a crappy miniseries anyways.

    For casual communications, there is e-mail.

    For vital ones, there is registered mail, fax or phone.

  13. DNSPark, too by mrmagos · · Score: 3, Interesting

    I use DNSPark, and they were subject to a DDOS attack earlier this week, too. Are they affiliated with EveryDNS too, or is it coincidence, since they are another cheap/free DNS host?

    --
    Never start vast projects with half-vast ideas.
  14. Re:Every DNS, not EasyDNS. by sirket · · Score: 3, Informative

    If your upstream provider can't handle 400Mbps of traffic then you're being hosted by a pretty shitty ISP/data-center. It's not like gig uplinks are expensive (even if you only commit to a tiny rate you can generally get gig uplinks). Spread this across 4 or more datacenters and you've got a lot of bandwidth.

    Not to mention that networking people generally don't give a shit about bandwidth- it's packets per second that kill routers, not bandwidth. Assuming 100 byte packets that's about 4Mpps- Even a basic 7600 can handle this kind of traffic. Assuming 30 byte packets (can't be smaller than that) you're talking about 15Mpps. Again Even a basic 7600 should be able to handle that- not to mention a Juniper M7i or similar. Most Foundry equipment would laugh at that rate. All of these routers can do ACL's at full packet rates.

    That said- other recent DNS attacks exceeded 1.5 Gigabits per second of traffic and were a lot more vicious than the attack being described here.

    I'm not knocking EveryDNS- I know what a bitch dealing with a DDoS can be- the problem tends to be that most people aren't ready to deal with it. Using BGP community based nullrouting most service can be restored within seconds of the target IP(s) being identified. That allows admins to keep untargeted systems and services up while the attacked systems are dealt with. The admins can then use the time to locate some/any pattern in the attack or enable the appropriate filtering such as a Cisco Riverguard or similar.

    -sirket

  15. Re:solution to DDOS attack by sirket · · Score: 5, Informative

    Not quite- It generally works like this:

    First off- be prepared for a damned attack and don't wait til it happens. When an attack does come:

    1- Identify the target IP address
    2- Immediately null-route traffic for that address (preferably using BGP community based null-routing)
    This gets the rest of your systems back up and gives you time to work on the problem.
    3- Try to identify a pattern in the attacking traffic- use a product from a company like Mazu- or just tcpdump if you're good with sed and awk.
    4- If there is a pattern ask the upstream ISP to block based on that pattern (same source port, same source IP, same TTL, whatever). Or block it yourself if you have the router and bandwidth capacity to deal with the attack yourself- though that's generally a waste of your resources.
    5- If there is no pattern but the traffic is malformed then enabled a Cisco Riverguard or similar protection device that can filter out malformed traffic at the higher protocol layers. As an alternative, sign up for such a service form a company like Prolexic.
    6- Remove your null route and see how you did.
    7- If you can't afford a protection service, you can try moving the host/dns records to new IP's. Sometimes the attacks don't follow- sometimes they do. It's often worth a try as it can be done faster than enabling protection services in many cases. In this case leave the old null route in place until the attack stops. Be prepared for the attack to return at any time once they realize what's happened.

    Make sure to keep traffic logs for law-enforcement and to share with other ISP's so that they can track down the offending bots.

    In the future try to keep your traffic as segregated as possible such that an attack on a single host will not take down too many other services should you need to null-route that address for an extended period of time.

    The easiest solution- block all IP addresses assigned to the APNIC region and watch as your site immediately returns to normal. Sadly most of the DDoS's I've seen recently had the majority of their traffic sourced from APNIC addresses.

    -sirket

  16. Re:Real ripple effects, even from this small event by Mixel · · Score: 3, Insightful
    DOS attacks are easy to pervent

    Learn to spell, get a clue.

    There is nothing you can practically do to prevent someone on the internet from sending a packet addressed to you, nor two packets, nor 1000000. There is nothing you can practically do to prevent the source address on each of those packets to be different. If a DOSer has much bigger pipes than you, you are sunk, unless you can do something very smart. For a start, getting remote access to your server during a DOS attack is tricky unless you have redundancy. Then you need to profile the traffic, find patterns which you can filter.

    The non-triviality of a (D)DOS is the reason why everyone is interested to learn how to defend against such attacks. This is why we want to hear how EveryDNS handled the problem so well. A second-rate admin would not be able to. While I appreciate your sentiment regarding "survival of the fittest". I feel it can be better expressed as "survival of the fittest admin for the job".

    And lets not try to discuss how, if they can carry this out, you are going to catch them
    As there are lots of admins on /., it is only natural that they want to get fitter through learning how to deal with such attacks. Therefore the how is very important and very much worth discussing. Problems don't magically resolve themselves, though it may seem like that because professional and dedicated people spend much of their time figuring out how to deal with them effectively.