Massive 300Gbps DDoS attack on media firm fuelled by unpatched server flaw

Supermicro IPMI flaw used to create huge botnet, says VeriSign

Hacktivists summoned up a massive and previously undocumented 300Gbps DDoS attack earlier this summer by exploiting an obscure motherboard-level flaw on 100,000 unpatched servers, VeriSign has revealed in its latest quarterly Trends report.

As with every other DDoS trend survey these days, VeriSign's analysis notes the growing size of DDoS attacks in the second quarter but the more startling news is what happened to one of its customers some time in the second quarter, possibly around June.

Described as being a content delivery network (CDN) in the media and entertainment sector, an unidentified datacentre found itself on the receiving end of a determined DDoS "siege", which started with a three-hour SYN and TCP flood, a standard softening up process.

After mitigation, the attackers changed tack not long after to use large UDP packets, quickly reaching a peak traffic volume of 250Gbps that required VeriSign to start shuffling the load around its global capacity. For the following 24 hours, the mitigation systems had to cope with more than 30 short but large bursts of UDP and TCP as the attackers probed for weaknesses.

In a final attempt to overwhelm defences the volume ramped up to 300Gbps, making this one of the largest DDoS attacks ever publically disclosed. Repelled by VeriSign's digital fortress, after 30 hours the attack ceased.

Ironically, the incredible size is only one of the significant aspects of this tale of DDoS woe.

Huge DDoS attacks of the 300Gbps ilk have exploited a number of configuration issues in order to generate their size, including the infamous DNS reflection attacks on Spamhaus in March 2013 and one exploiting the NTP protocol against CloudFlare earlier in 2014.

The DDoS attack recorded by VeriSign summoned its immense power from a botnet made up of as many as 100,000 servers vulnerable to the 'Supermicro IPMI [Intelligent Platform Management Interface]' flaw, made public by researcher Zachary Wikholm on 19 June. This motherboard-level issue allowed hackers to gain access to an unencrypted password file for a system by connecting through software port 49152.

It appears to have been that simple - the attacks could automate the harvesting of passwords from a distributed web of apparently legitimate servers around the world, allowing these to be used at the botnet's bidding.

Worries about flaws in the widely-used Supermicro boards and their IPMI management system used by many big-brand server vendors, are not new. In late 2013, security firm Rapid7 warned of serious issues with the software, only weeks after the University of Michigan researchers carried out tests that showed 40,000 servers were vulnerable to similar flaws.

Supermicro apparently told Wikholm that it had patched the reported vulnerability as long ago as 2013. What is clear is that the industry knew for some time there were problems with the interface but large numbers of servers seem not to have been patched, possibly because doing this involves the inconvenience of shutting down each affected system.

The results of this are now clear - attackers noticed the issue and used it in a real-world attack.

"The 300Gbps attack that we mitigated in Q2 was the third largest known attack, but what was most surprising about it was that reflective amplification techniques were not used in the attack. It was straight bot traffic. This is very rare for an attack of this size," VeriSign CSO, Danny McPherson, told Techworld.

As to who was behind the super-bot, as with the previous two massive DDoS attacks on Spamhaus and CloudFlare he suggested hacktivism or ideology as the likely mojo.

"The motivations for the attacks appear to stem from recent controversies around server monetization and potentially other disputes around moderation on independent forums," was as far as McPherson would go. Could this be code for the politics of Bitcoins? That's a guess but only a guess.

The attackers knew what they were doing. Another apparent technique was to sneak malformed packets past mitigation by putting them inside GRE (Generic Routing Encapsulaiton) tunnels.

VeriSign's report also found that the media and entertainment industry accounted for 43 percent of all DDoS attacks in Q2, just ahead of IT services, cloud firms and SaaS providers on 41 percent.

Victims and mitigation firms rarely talk about this kind of attack but perhaps Verisign's frankness will unfreeze that inhibition. The world needs insight into real DDoS attacks and not simply more abstract discussion about rising traffic volumes. On this occasion we have a simple punchline to thank for news of this attack - VeriSign won and the DDoS attackers lost.