Attackers leveled a massive distributed denial-of-service attack against a specific website in early June, topping a bandwidth of 1.44 terabits-per-second and 385 million packets-per-second, the largest volumetric attack encountered by Internet infrastructure firm Akamai.
Akamai said a second attack topped 500 gigabits-per-second and targeted a different site at the same provider. While the company would not give more details about the targeted sites, Akamai believes the attacker focused on the websites for a "social" reason, and the hosting provider was not the actual target, says Roger Barranco, vice president of global security operations for Akamai.
The sheer variety of data types used in the attack, along with signs of significant coordination, suggests a knowledgeable attacker, he says.
"What was really different is that the absolute kitchen sink was thrown in for this attack," Barranco says. More than nine different types of traffic were used in the attack - far more than the two or three types used in a typical attack - and the traffic surge lasted about an hour. Most attacks are measured in minutes, so this was notable. "It's has been a long time that we've seen the duration that long and an attack of that size."
A typical year of denial-of-service attacks includes a massive number of smaller attacks targeting gaming sites, often to give one player an edge over the competition, and a few massive bandwidth attacks. Over the past 24 months, the median denial-of-service attack has peaked at less than 250,000 packets-per-second, according to Akamai's State of the Internet (SOTI) report. Very few attacks have surged far beyond that average. Security firm Imperva recorded a 500 million packet-per-second attack in early 2019, for example.
The attack weathered by Akamai and its customer came close to that packet rate, but surpassed the bandwith of that previous attack because the average packet seen by Akamai consisted of more data. The attackers used volumetric attacks, or floods, of ACK, SYN, UDP, NTP, TCP reset, and SSDP packets. In addition, other attacks included CLDAP reflection, TCP anomaly, UDP fragments, and multiple botnet attack tools, according to Akamai.
The variety of attack types, along with the coordination among the tools, was a novel approach, Barranco says.
"Really, you have to go all the way back to the Mirai days, to find a time when there was something that we hadn't seen before and we had to really engineer something unique to handle it," he says.
The attack used traffic from sources worldwide, many more than the denial-of-service attacks created by the Mirai botnet, Barranco says. The largest Mirai attack topped out at more than 600 megabits-per-second.
"What was really different in this case was the coordination," he says. "The actors were able to get a hold of multiple tools. It did not source from a single region. So that means that someone went way out of their way to either reserve the capability or collect the tools needed to level an attack of this size."
Akamai noted that the company quickly adjusted to the massive denial-of-service attack. The bulk of the attack was handled in seconds, but it took about ten minutes to get it 100% under control, says Barranco.
He recommends that companies know their traffic patterns and what traffic is necessary for business before they are targeted by an attack. "A lot of enterprises do not have a good handle on their traffic patterns," he says. "If you get hit with a DDoS, you have a good chance of being able to mitigate it, but you also have a good chance of mitigating good traffic."
With remote working becoming more common during the coronavirus epidemic, revisiting the assumptions about good sources of traffic can help a company get ready for a DDoS attack.