How Internet Escaped Being Shut Down by DNS Attack

Last Friday was a tough day for the internet, though it and its inherently redundant design pulled through. Early in the morning on Oct. 21, multiple big-name sites, including Twitter, Spotify and GitHub, experienced slowdowns as an attack against Domain Name System (DNS) services took shape.

The initial attack was focused against DNS provider DynDNS, starting at approximately 7:10 a.m. EDT and resolved by 9:20 a.m. EDT. DynDNS was subsequently hit by a second attack at 11:52 a.m. EDT, which was mitigated by 1 p.m. EDT. A third attack wave also hit DynDNS later in the day, though the company said it was mitigated without any impact.

“At this point we know this was a sophisticated, highly distributed attack involving 10s of millions of IP addresses,” Kyle York, chief strategy officer at Dyn, wrote in a blog post. “We can confirm, with the help of analysis from Flashpoint and Akamai, that one source of the traffic for the attacks were devices infected by the Mirai botnet.”

The Mirai botnet is a collection of infected internet of things (IoT) devices, including cameras as well as routers, that is behind the largest distributed denial-of-service (DDoS) attack ever publicly reported. Mirai has been implicated in the 1T-bps attack against internet service provider OVH as well as a 665G-bps attack against security blogger Brian Krebs. Mirai compromises devices that are using known or default passwords.

While the attack against DynDNS impacted users on the East Coast of the United States, it did not shut down the internet. In fact, the attack didn’t even shut down all of DynDNS.

“We should note that Dyn did not experience a system-wide outage at any time—for example, users accessing these sites on the West Coast would have been successful,” York noted.

The DNS is a highly distributed system that is globally available through a network of root, authoritative and recursive nameservers. When DynDNS first started to experience problems, it was possible for affected websites to change their nameservers to ones that weren’t under attack. For example, Amazon Web Services (AWS) customers could simply choose to rely on Amazon’s route 53 for DNS. While DNS doesn’t instantly update around the world—meaning there would be some intermittent availability issues during a DNS nameserver change—one DNS service under attack doesn’t mean the whole internet is down.

Most websites will choose to set at least two nameservers for each domain record for a given website. The basic idea behind having two or more nameservers is for resiliency and availability, though when nameservers are at the same service and are under attack, there can still be an issue.

On Oct. 21, one thing that many of the websites affected by the attack were doing was changing nameservers. The DNSstream service managed by Cisco’s OpenDNS business unit showed a steady stream of nameserver updates on Oct. 21, as organizations attempted to adjust.

In many ways, the internet attack is a wakeup call for organizations to configure DNS for optimal resiliency. More specifically, that means using two (or more) DNS providers and listing multiple nameservers for added resiliency. It’s also yet another wakeup call for IoT security as the risk of default passwords and unsecured devices is no longer a theoretical one.

Sean Michael Kerner is a senior editor at eWEEK and InternetNews.com. Follow him on Twitter @TechJournalist