Application layer DDoS attacks are becoming more common, perhaps because they cost less for malicious actors to execute and can more effectively evade defenses than network layer attacks, Imperva says.
One such attack generated 8.7Gbps at its peak, “unheard of in relation to application layer assaults,” in an effort to thread its way through the DDoS mitigations that had been set up to defend against such attacks, according to Imperva’s “DDoS Threat Landscape Report 2015-2016” released today.
Application layer attacks can be measured in responses per second required from the application targeted, and they generally require less volume than network layer attacks to succeed, the report says. That means they require fewer botnet resources, if botnets are the platform from which the attack is launched.
That in turn means attackers on a budget can afford to create attacks that last longer. More than 44 percent of application layer attacks lasted more than an hour while just 12.2 percent of network layer ones did.
Imperva gathered the information from actual attacks against customers of its Incapsula cloud-based application delivery and security service. The report covers the period from April 1, 2015 to March 31, 2016.
Over the period of the report 60 percent of attacks were at the application layer and 40 percent at the network layer, although if trends continue, the split could be 50-50 by 2018. The network attacks are able to generate faster packet-forwarding rates because network packets are smaller and able to create higher peak DDoS traffic, according to the report.
Combined, both application and network layer attacks averaged 445 per week, with the number per week increasing over the course of the year.
The number is up in part because it’s easier to become a DDoS attacker given the rise of DDoS for hire services masquerading as stressor tools legitimate enterprises can use to test their own defenses, Imperva says. Over the period of the report the percentage of attacks using DDoS-for-hire rose from 63.8 percent to 93 percent. “These are non-professionals who use DDoS for racketeering or to instigate attacks out of boredom or spite,” the report says.
Imperva says part of the jump is because its Incapsula service has gained more customers over that year, so there is a bigger base sample of users. Many of those new customers sign on because they’re having DDoS troubles, either actual attacks or threatened attacks in which attackers are trying to extort money in exchange for not attacking.
Another tactic that is becoming more common is launching lower volume attacks one after another after another. Consecutive attacks that aren’t devastating in themselves, but they work to exhaust mitigation teams, making them less effective and so can be successful in degrading service. They are also used as decoys to deflect attention against other types of attacks going on at the same time, the report says.
The peak size of attacks is also increasing, with the biggest Imperva has seen – 470Gbps at its height – hitting during the second quarter of this year, which is after the period of the report. These massive attacks rely on the size of botnets that are used to launch them. That particular attack was at the network layer.
This finding goes along with a similar report earlier this year by Akamai.
The vast majority – 87.8 percent of all attacks – lasted less than 30 minutes, today's Imperva report says. The longest lasted 54 days.
Most attacks were carried out on a single attack vector, but some, in the low single-digits percentages, were carried out with up five. The most popular attacks were UDP floods, then SYN floods, TCP floods and Large SYN floods.
Over the course of the report China was the number one source of DDoS attacks, and the U.S. was where most of the DDoS attacks landed.