BotBeat
...
← Back

> ▌

Independent ResearchIndependent Research
RESEARCHIndependent Research2026-03-02

Comprehensive Study Reveals 43,000 Blocked Domains Across Indian ISPs, Exposing Massive DNS Censorship Scale

Key Takeaways

  • ▸The study identified 43,083 blocked domains across six major Indian ISPs—the largest known blocklist for India—by testing 294 million domains rather than relying on curated lists
  • ▸Research methodology involved 1.76 billion DNS queries, detecting blocking through unique signatures including forged IPs and DNS injection techniques specific to each ISP
  • ▸Findings revealed previously undocumented overblocking caused by misconfigurations, suggesting censorship infrastructure affects more domains than intentionally targeted
Source:
Hacker Newshttps://dnsblocks.in/#explorer↗

Summary

A groundbreaking research study titled 'Poisoned Wells' has uncovered the largest known list of blocked domains in India, identifying 43,083 censored websites across six major Indian Internet Service Providers. Conducted by researcher Karan Saini with support from the Open Technology Fund and the Internet Governance Project at Georgia Institute of Technology, the study represents the most comprehensive examination of DNS censorship in India to date. Unlike previous research that relied on curated lists of potentially blocked websites, this study tested the entire visible domain name space—294 million registered domains—across ISPs including Reliance Jio, Bharti Airtel, ACT, MTNL, Connect Broadband, and You Broadband.

The research methodology employed sophisticated DNS measurement techniques, deploying direct ISP connections and identifying infrastructural resolvers to query domains at rates designed to avoid service degradation. The study identified unique blocking signatures for each ISP, including forged IP addresses and DNS injection techniques. Researchers validated suspected blocks through control measurements using third-party DNS providers and categorized the blocklist using an expanded version of Citizen Lab's taxonomy, combining open-source lists, manual classification, and machine learning approaches.

The findings reveal not only intentional blocking but also previously undocumented instances of overblocking caused by misconfigurations, suggesting that censorship infrastructure in India affects far more domains than officially targeted. The complete dataset—comprising 1.76 billion DNS queries totaling approximately 760 GB of raw data—has been released under a Creative Commons license, along with an interactive explorer tool that allows researchers and the public to examine blocking patterns across ISPs, categories, and top-level domains. The study represents a significant contribution to internet freedom research and transparency around digital censorship practices in one of the world's largest internet markets.

  • The complete dataset, raw measurements, and interactive explorer have been released publicly under Creative Commons license to support transparency and further research
  • The study was supported by the Open Technology Fund and hosted by the Internet Governance Project at Georgia Institute of Technology

Editorial Opinion

This research represents a methodological leap forward in internet censorship studies by examining the entire visible domain space rather than testing pre-selected lists, revealing the true scale of DNS blocking in India. The discovery of significant overblocking due to misconfigurations raises serious concerns about collateral censorship affecting legitimate websites and services. The researchers' decision to release the complete dataset publicly—despite its massive size—demonstrates a commitment to transparency that should set the standard for future internet freedom research, enabling independent verification and broader analysis of digital rights issues in democratic societies.

Data Science & AnalyticsGovernment & DefenseRegulation & PolicyPrivacy & DataMisinformation & Deepfakes

More from Independent Research

Independent ResearchIndependent Research
RESEARCH

How AI Discourse in Training Data Shapes Model Alignment, Study Shows

2026-05-18
Independent ResearchIndependent Research
RESEARCH

Distribution Fine Tuning: New Algorithm Eliminates LLM 'Slop' and Boosts Creativity 164%

2026-05-18
Independent ResearchIndependent Research
RESEARCH

MemEye Framework Reveals Gaps in Multimodal Agent Memory: Current VLMs Struggle with Fine-Grained Visual Details

2026-05-18

Comments

Suggested

AnthropicAnthropic
POLICY & REGULATION

Advanced AI Models Bring Government to 'Reflection Point,' CIA Official Says

2026-05-20
OpenAIOpenAI
FUNDING & BUSINESS

OpenAI Prepares for IPO After Musk Lawsuit Threat Clears

2026-05-20
Google / AlphabetGoogle / Alphabet
PARTNERSHIP

Singapore Inks AI Deals with Google

2026-05-20
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us