Threat actors specializing in distributed denial-of-service (DDoS) attacks have started abusing network middleboxes for reflection and amplification, Akamai warns.
The use of misconfigured network middleboxes and censorship systems for DDoS reflection was theorized last year by a group of researchers at the University of Maryland and University of Colorado Boulder.
In their paper, the academics showed that censorship infrastructure could be abused to achieve DDoS amplification ratios of up to 700,000:1. Furthermore, they showed that firewalls and intrusion prevention systems deployed within non-censoring nation-states could also be weaponized.
Even if no valid TCP connection or handshake has been established, middleboxes may respond to requests with very large block pages.
In a report today, Akamai explains that attackers can craft TCP packet sequences and send them to middleboxes. If the HTTP request headers of these sequences contain a domain name for a blocked site, the middlebox responds with HTTP headers or entire HTML pages.
As part of a DDoS attack, the adversary spoofs the source IPs of the intended victim, which results in the middleboxes directing traffic to that IP.
“These responses provide attackers with a reflection opportunity, and in some cases a significant amplification factor,” Akamai notes.
Although still small compared to other vectors, attacks that abuse the “TCP Middlebox Reflection” technique appear to be growing in popularity, Akamai says. To date, the method has been used against the banking, gaming, media, travel, and web-hosting sectors.
While the first attacks only peaked at 50Mbps, the most recent assaults hit 2.7 gigabytes per second (Gbps) and 11 Gbps, the latter also peaking at 1.5 million packets per second (Mpps).
With hundreds of thousands of middlebox systems worldwide vulnerable to such attacks, an adversary no longer needs access to a large number of compromised systems and the potential for TCP reflection abuse is very high, especially since TCP Middlebox Reflection has been tested and tried.
Mitigation options, however, are relatively easy to implement, Akamai says. Because SYN packets are typically used to initiate the TCP handshake and not for data transmission, any such packet that has a length greater than 0 bytes is suspect and can be used to trigger defenses.
“SYN challenges may also be effective at preventing middlebox resource exhaustion effects. The middlebox will not properly handle the resulting challenge packet, so the SYN packets won’t make it past mitigation gear, and since the handshake will never complete, data flows should also be dropped before making it to servers and applications,” Akamai says.