DAmpADF: A framework for DNS amplification attack defense based on Bloom filters and NAmpKeeper
Yunwei Dai,Tao Huang,Shuo Wang
DOI: https://doi.org/10.1016/j.cose.2024.103718
IF: 5.105
2024-01-24
Computers & Security
Abstract:Domain Name System (DNS) amplification attacks exploit botnets and open recursive DNS servers to launch Distributed Denial of Service (DDoS) attacks. During an attack, the attacker leverages infected computers (bots) to perpetually send small spoofed DNS queries to numerous open recursive DNS servers. The servers, in turn, respond with lots of large DNS responses, which are reflected back to the victim. Such responses are usually several times larger than the original queries, and could exhaust the resources of CPUs, memory, and network bandwidth, rendering them unavailable for benign users. However, most in-network DDoS mitigation systems today inevitably cause normal DNS responses to be discarded while scrubbing traffic, as they do not distinguish between legitimate and malicious responses. To address this issue, some existing solutions employ Bloom filters to filter out unsolicited DNS responses, utilizing the "one-to-one mapping" relationship between DNS queries and responses. In this work, we present a framework called DAmpADF, designed to defend against DNS amplification attacks. The framework employs Bloom filters at the edge or core routers of Internet Service Providers (ISPs) or organizations to filter out malicious DNS responses. To reduce the false positives of the Bloom filters, we propose a novel data structure called the Non-amplifier Keeper (NAmpKeeper), which maintains the most frequently queried DNS servers that are not DNS amplifiers. By excluding queries to non-amplifiers from the Bloom filters, the false positives of Bloom filters are decreased significantly. Experimental results show that DAmpADF outperforms previous methods and achieves a superior filtration ratio of illegitimate DNS responses. Furthermore, the proposed approach incurs small, constant processing and memory overhead, enabling support for high line rates.
computer science, information systems