In a security alert sent to its customers and shared with ZDNet this week, Radware said that during the last week of 2020 and the first week of 2021, its customers received a new wave of DDoS extortion emails.And Dan Goodin reports on the latest technique the DDOS-ers are using in DDoSers are abusing Microsoft RDP to make attacks more powerful:
Extortionists threatened companies with crippling DDoS attacks unless they got paid between 5 and 10 bitcoins ($150,000 to $300,000)
The security firm believes that the rise in the Bitcoin-to-USD price has led to some groups returning to or re-prioritizing DDoS extortion schemes.
As is typical with many authenticated systems, RDP responds to login requests with a much longer sequence of bits that establish a connection between the two parties. So-called booter/stresser services, which for a fee will bombard Internet addresses with enough data to take them offline, have recently embraced RDP as a means to amplify their attacks, security firm Netscout said.I don't know why it took me so long to figure it out, but reading Goodin's post I suddenly realized that techniques we described in Impeding attrition attacks in p2p systems, a 2004 follow-up to our award-winning 2003 SOSP paper on the architecture of the LOCKSS system, can be applied to preventing systems from being abused by DDOS-ers. Below the fold, brief details.
The amplification allows attackers with only modest resources to strengthen the size of the data they direct at targets. The technique works by bouncing a relatively small amount of data at the amplifying service, which in turn reflects a much larger amount of data at the final target. With an amplification factor of 85.9 to 1, 10 gigabytes-per-second of requests directed at an RDP server will deliver roughly 860Gbps to the target.
Among the LOCKSS system's defenses against abuse are two relevant to DDOS prevention, rate limits and effort balancing.
Rate LimitsI've written before about the importance of rate limits, quoting Paul Vixie:
Every reflection-friendly protocol mentioned in this article is going to have to learn rate limiting. This includes the initial TCP three-way handshake, ICMP, and every UDP-based protocol. In rare instances it's possible to limit one's participation in DDoS reflection and/or amplification with a firewall, but most firewalls are either stateless themselves, or their statefulness is so weak that it can be attacked separately. The more common case will be like DNS [Response Rate Limiting], where deep knowledge of the protocol is necessary for a correctly engineered rate-limiting solution applicable to the protocol.The RDP server being used to DDOS sees a flood of authentication requests whose source address has been spoofed to be the target of the DDOS. This isn't what they'd see from a real user, so the RDP server should rate-limit sending authentication responses to a client to a reasonable rate for a real client. This would be helpful, but it isn't enough. Because the DDOS-ers use a large number of systems to mount an attack, even a fairly low rate of reponses can be harmful.
Effort BalancingIn our 2004 paper we wrote:
Effort Balancing. If the effort needed by a requester to procure a service from a supplier is less than the effort needed by the supplier to furnish the requested service, then the system can be vulnerable to an attrition attack that consists simply of large numbers of ostensibly valid service requests. We can use provable effort mechanisms such as Memory-Bound Functions to inflate the cost of relatively “cheap” protocol operations by an adjustable amount of provably performed but otherwise useless effort. By requiring that at each stage of a multi-step protocol exchange the requester has invested more effort in the exchange than the supplier, we raise the cost of an attrition strategy that defects part-way through the exchange. This effort balancing is applicable not only to consumed resources such as computations performed, memory bandwidth used or storage occupied, but also to resource commitments. For example, if an adversary peer issues a cheap request for service and then defects, he can cause the supplier to commit resources that are not actually used and are only released after a timeout (e.g., SYN floods). The size of the provable effort required in a resource reservation request should reflect the amount of effort that could be performed by the supplier with the resources reserved for the request.Vixie also noted the economic requirement:
Engineering economics requires that the cost in CPU, memory bandwidth, and memory storage of any new state added for rate limiting be insignificant compared with an attacker's effort.The reason RDP can be used to amplify a DDOS attack is that, as Goodin wrote:
RDP responds to login requests with a much longer sequence of bits that establish a connection between the two parties.The obvious application of effort balancing would be to require that RDP's login requests be padded with additional bytes to make them longer than the login reponse. Thus the RDP server would act to attenuate the attack, not amplify it. This would satisfy Vixie's goal:
Attenuation also has to be a first-order goal—we must make it more attractive for attackers to send their packets directly to their victims than to bounce them off a DDoS attenuator.The protocol could specify that the padding bytes not be random, but be computed from the login request parameters by some algorithm making them relatively expensive to generate but cheap to verify (cf. proof-of-work). This would not significantly impact legitimate clients, who issue login requests infrequently, but would increase the cost of using the RDP server to disguise the source of the attack.