Biblio

Load balancing and IP anycast are traffic routing algorithms used to speed up delivery of the Domain Name System. In case of a DDoS attack or an overload condition, the value of these protocols is critical, as they can provide intrinsic DDoS mitigation with the failover alternatives. In this paper, we present a methodology for predicting the next DNS response in the light of a potential redirection to less busy servers, in order to mitigate the size of the attack. Our experiments were conducted using data from the Nov. 2015 attack of the Root DNS servers and Logistic Regression, k-Nearest Neighbors, Support Vector Machines and Random Forest as our primary classifiers. The models were able to successfully predict up to 83% of responses for Root Letters that operated on a small number of sites and consequently suffered the most during the attacks. On the other hand, regarding DNS requests coming from more distributed Root servers, the models demonstrated lower accuracy. Our analysis showed a correlation between the True Positive Rate metric and the number of sites, as well as a clear need for intelligent management of traffic in load balancing practices.

Distributed Denial-of-Service (DDoS) attacks continue to be a major threat on the Internet today. DDoS attacks overwhelm target services with requests or other traffic, causing requests from legitimate users to be shut out. A common defense against DDoS is to replicate a service in multiple physical locations/sites. If all sites announce a common prefix, BGP will associate users around the Internet with a nearby site, defining the catchment of that site. Anycast defends against DDoS both by increasing aggregate capacity across many sites, and allowing each site's catchment to contain attack traffic, leaving other sites unaffected. IP anycast is widely used by commercial CDNs and for essential infrastructure such as DNS, but there is little evaluation of anycast under stress. This paper provides the first evaluation of several IP anycast services under stress with public data. Our subject is the Internet's Root Domain Name Service, made up of 13 independently designed services ("letters", 11 with IP anycast) running at more than 500 sites. Many of these services were stressed by sustained traffic at 100× normal load on Nov. 30 and Dec. 1, 2015. We use public data for most of our analysis to examine how different services respond to stress, and identify two policies: sites may absorb attack traffic, containing the damage but reducing service to some users, or they may withdraw routes to shift both good and bad traffic to other sites. We study how these deployment policies resulted in different levels of service to different users during the events. We also show evidence of collateral damage on other services located near the attacks.