Mind the IP Gap Measuring the impact of IPv6 on DNS censorship

Ian Martiny, Hammas Bin Tanveer, Jack Wampler, Rishab Nithyanand, Eric Wustrow · 2025

Read Mind the IP Gap Measuring the impact of IPv6 on DNS censorship by Ian Martiny, Hammas Bin Tanveer, Jack Wampler, Rishab Nithyanand, Eric Wustrow (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-07197.txt.

Introduction

Ian Martiny and co-authors measure internet censorship of the newer IPv6 network across the world, in a paper from 2025. Most past measurement work used the older IPv4 network. As IPv6 spreads, it was unclear whether censors block it as well as they block IPv4.

The team found resolvers that serve both network versions and sent them more than twenty million name requests. Blocking was measured at resolver, network, and country levels, and the traits of blocked domains were studied. Nearly all censors did block on IPv6, but their rules were uneven and often weaker than on IPv4. The authors argue that censors must spend more to reach parity. For now, the gap offers new routes for circumvention research.

Catalog record

CREATORIan Martiny, Hammas Bin Tanveer, Jack Wampler, Rishab Nithyanand, Eric Wustrow
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE87 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-07197

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Mind the IP Gap Measuring the impact of IPv6 on DNS censorship Ian Martiny, Hammas Bin Tanveer, Jack Wampler, Rishab Nithyanand, Eric Wustrow Abstract Internet censorship impacts large segments of the Internet, but so far, prior work has focused almost exclusively on performing measurements using IPv4. As the Internet grows, and more users connect, IPv6 is increasingly supported and available to users and servers alike. But despite this steady growth, it remains unclear if the information control systems that implement censorship (firewalls, deep packet inspection, DNS injection, etc) are as effective with IPv6 traffic as they are with IPv4. In this paper, we perform the first global measurement of DNS censorship on the IPv6 Internet. Leveraging a recent technique that allows us to discover IPv6-capable open resolvers (along with their corresponding IPv4 address), we send over 20 million A and AAAA DNS requests to DNS resolvers worldwide, and measure the rate at which they block, at the resolver, network, and country level as well examine the characteristics of blocked domains. We observe that while nearly all censors support blocking IPv6, their policies are inconsistent with and frequently less effective than their IPv4 censorship infrastructure. Our results suggest that supporting IPv6 censorship is not all-or-nothing: many censors support it, but poorly. As a result, these censors may have to expend additional resources to bring IPv6 censorship up to parity with IPv4. In the meantime, this affords censorship circumvention researchers a new opportunity to exploit these differences to evade detection and blocking. 1 Introduction Internet censorship is a global problem that affects over half the world’s population. Censors rely on sophisticated network middleboxes to inspect and block traffic. A core component of Internet censorship is DNS blocking, and prior work has extensively studied how DNS censorship occurs, both for specific countries [ 6 , 22 ] and globally [ 24 , 14 , 33 , 36 ] . These studies generally perform active measurements of DNS resolvers for large sets of domains, and identify forged censorship responses from legitimate ones. Unfortunately, this prior work has focused exclusively on the IPv4 Internet, in part because scanning the IPv6 Internet for open resolvers is difficult [ 28 ] , owing to its impossible-to-enumerate 128-bit address space. In this paper, we perform the first comprehensive global measurement of DNS censorship on the IPv6 Internet. We leverage a recent network measurement technique that can discover dual-stack IPv6 open resolvers from their IPv4 counterpart [ 21 ] , and use these IPv4-IPv6 resolver pairs to study DNS censorship globally. We then use this data to measure the difference in censorship on IPv4 and IPv6. IPv6 is becoming more widely deployed, with nearly 35% of current Internet traffic being served over native IPv6 connections [ 4 ] . However, IPv6 has fundamentally different performance characteristics [ 15 ] , network security policies [ 13 ] , and network topologies [ 12 ] compared to the traditional IPv4 Internet. While it may seem that censors either do or don’t support detecting and censoring IPv6 DNS in an all-or-nothing fashion, we find that there is a tremendous range of how well a censor blocks in IPv6 compared to IPv4. In particular, although nearly all of the countries we study have some support for IPv6 censorship, we find that most block less effectively in IPv6 compared to IPv4. For instance, we observe Thailand censors on average 80% fewer IPv6 DNS resources compared to IPv4 ones, despite a robust nation-wide censorship system [ 19 ] . Studying censorship in IPv6 can provide opportunities for circumvention tools. By identifying ways that censors miss or incorrectly implement blocking, we can offer these as techniques that tools can exploit. Moreover, because of the complex and heterogeneous censorship systems censors operate, many of these techniques would be costly for censors to prevent, requiring investing significant resources to close the IPv4/IPv6 gap in their networks. For this reason, we believe IPv6 can provide unique techniques for circumvention researchers and tool developers alike, that will be beneficial in the short term and potentially robust in the longer term. We find a significant global presence of IPv6 DNS censorship — comparable, but not identical to well documented IPv4 censorship efforts. Censors demonstrate a clear bias towards IPv4, censoring A queries in IPv4 at the highest rates, and a propensity for censoring native record types ( A in IPv4, AAAA in IPv6). At the country level we break down differences by resolver and domain across resource record and interface type. We find that multiple countries — Thailand, Myanmar, Bangladesh, Pakistan, and Iran — present consistent discrepancies across all resolvers or domains indicating centrally coordinated censorship, where the policies that govern IPv4 and IPv6 censorship are managed centrally. Other countries show more varied discrepancies in the ways that resolvers censor IPv4 and IPv6, due to decentralized models of censorship, such as that in Russia [ 34 ] , or due to independent and varied corporate network firewalls. We also identify behavior indicative of censorship oversight that can be advantageous to censorship circumvention. For example Brazil and Thailand censor IPv6 queries that rely on 6to4 bridges at lower rates, presumably due to the encapsulation of an IPv6 DNS request in an IPv4 packet, instead of appearing as UDP. Taken all together, this study provides a first look at IPv6 DNS censorship and the policy gaps that arise from the IPv6 transition. We provide the following contributions: • We conduct the first large-scale measurement of IPv6 DNS censorship in over 100 IPv6-connected countries. We find that while most censors support IPv6 in some capacity, there are significant gaps in how well they censor IPv6. • We provide methodological improvements on measuring DNS censorship that avoids relying on cumbersome IP comparisons (that are not robust to region-specific DNS nameservers). Our methods are easily reproducible, and can be used in future measurement studies. • We characterize the difference in censorship of both network type (IPv4 and IPv6), and resource type ( A and AAAA record), and identify trends in several countries. • Using our findings, we suggest several new avenues of future exploration for censorship circumvention researchers, and censorship measurements. The remainder of this paper is organized as follows. Section 2 provides background information in DNS censorship, and the relation of IPv6 to relevant DNS infrastructure. We outline our compiled methodology and ethical design considerations in Section 3 before presenting our findings on the global prevalence of IPv6 censorship in Section 4 . We then dig into per country analysis based on Resource Record types in Section 5 and IP protocol version in Section 6 . We select several case studies to highlight in Section 7 before covering related work in Section 8 . Finally Section 9 provides discussion and contextualization of this work before concluding. 2 Background Our work is focused on uncovering differences in Internet censorship that occur from DNS censorship mechanisms’ failure to effectively adapt to the Internet’s adoption of the IPv6 protocol . In this section, we provide an overview of Internet censorship approaches ( Section 2.1 ), DNS censorship mechanisms and infrastructure ( Section 2.2 ), and the impact of IPv6 on DNS censorship ( Section 2.3 ). 2.1 Internet

Read the complete text