A group of security researchers has identified a new distributed denial-of-service (DDoS) attack vector over TCP, which allows for reflected amplification at previously unseen levels, with amplification ratios of up to 700,000:1.
The HTTP-based reflected amplification attack leverages misconfigured network middleboxes and censorship systems, including devices that allow for infinite amplification, as they respond to an attacker’s packet with an endless stream of packets.
“These attacks can produce orders of magnitude more amplification than existing UDP-based attacks,” researchers from the University of Maryland (Kevin Bock, Dave Levin, Yair Fax, and Kyle Hurley) and University of Colorado Boulder (Abdulrahman Alaraj and Eric Wustrow) say.
“Collectively, our results show that censorship infrastructure poses a greater threat to the broader Internet than previously understood,” the researchers add.
The identified technique, they explain, also allows for the weaponization of firewalls and intrusion prevention systems that have been deployed within non-censoring nation-states.
Unlike most DDoS amplification techniques today, which are UDP-based, the new technique uses TCP, despite the 3-way handshake that protects TCP applications from being abused as amplifiers. The reason, the researchers explain, is that there are network middleboxes that “do not conform to the TCP standard,” thus opening the door to abuse.
Specifically, the researchers discovered a large number of censorship middleboxes that would respond to censored requests with large block pages, although no valid TCP connection or handshake exists. Thus, these devices can be weaponized and abused for DDoS amplification.
The academics used in their research an experimental genetic algorithm for evading censorship named Geneva (Genetic Evasion), which allowed them to identify misconfigured middleboxes.
Working with 184 sample middleboxes designed for HTTP censorship by injecting block pages, the researchers discovered 5 packet sequences that could be used to obtain amplified responses, along with another 5 modifications that could be used to further increase amplification for some of the middleboxes.
The researchers found several types of attacks, including normal TCP reflection, middlebox reflection, combined destination and middlebox reflection, routing loop reflection, and victim-sustained reflection. They also discovered that infinite amplification was caused by routing loops and victim-sustained reflection.
“Recall that we are searching for weaknesses in the TCP implementation in middleboxes, not in the TCP protocol itself. In addition, each middlebox has its own injection policies and block pages: this means that there is no one single amplification factor for this attack, since each middlebox we trigger will be different,” the researchers point out.
The researchers also note that they identified a set of IP addresses that act as mega-amplifiers, allowing for amplification factors of between 1,000,000 and 100,000,000. Furthermore, they say, some of these lead to infinite amplification factors, where some of the amplifiers would send an endless stream of packets, sometimes for days, even at full bandwidth.
They also point out that, because nation-states’ censorship infrastructure typically processes all of the traffic entering or exiting the country, all IP addresses behind an abused middlebox may appear as the source IP address of the traffic sent to the victim during a DDoS attack.
“Collectively, our results show that there is significant, untapped potential for TCP-based reflective amplification attacks,” the academics say, explaining that the true capacity available to an attacker using this attack vector is yet unknown.
The researchers shared their findings with several country-level CERTs (Computer Emergency Readiness Team), as well as with DDoS mitigation services and firewall manufacturers, but noted that mitigating the attack would require investment in changes that could weaken censorship infrastructure.
“Most nation-state censorship infrastructure is currently vulnerable, as well as many off-the-shelf commercial firewalls,” the researchers say. “Unfortunately, we found that this attack can [be] directed at practically anyone, whether [they] live inside a censored regime or not.”
According to application and network performance management company Netscout, these types of attacks have yet to be seen in the wild.
Related: Cloudflare Mitigated Record-Setting 17.2 Million RPS DDoS Attack
Related: Several DDoS Attack Records Broken in 2020