2020-anonymous-triplet-censors

Triplet Censors: Demystifying Great Firewall's DNS Censorship Behaviorcore

Abstract

We analyze the GFW's DNS injection behavior over nine months using the Alexa top 1M domains. Three distinct DNS injectors are fingerprinted by IP-DF/IP-TTL/DNS-AA/DNS-TTL combinations; one injector echoes the probe IP TTL, with implications for TTL-limited probing. We also observe groups of public IPs used to filter specific sets of domains.

Team notes

The three-injector model from this paper is foundational for any DNS- based circumvention reasoning. The TTL-mirroring injector specifically is a useful adversarial fingerprint: a measurement that observes it is almost certainly observing the GFW. For Lantern, the takeaway is that DNS-based bootstrap channels (DoH, DNSTT, kindling's DNS path) must assume the GFW is forging responses for many more domains than the obvious blocklists suggest.

Tags

censors
cn
techniques
dns-poisoningdpi
method
measurement-study