Volunteer probes outperform commercial VPN data for censorship research

Our team at Net Neutrality Monitor relies on distributed volunteer probes because they capture ground-truth network behaviour that commercial VPN telemetry often smooths over. When a probe in a Brisbane suburb or a Perth metro area runs a test, the result reflects what an ordinary household router actually sees, not what a data centre endpoint reports back. This distinction matters when identifying ISP-level filtering.

Commercial VPN services sell anonymity, speed, and geo-unblocking. Their datasets measure their own fleet's performance, route latency, and server availability. They rarely publish raw packet captures or DNS query logs from end-user devices. The methodology points to a different question: how well does the VPN tunnel work, not how much filtering exists between an Australian home connection and the wider internet.

Volunteer probes invert that priority. A probe is a lightweight script that resolves domain names, opens TLS connections, and records the responses it gets. In Australia, where the online landscape includes Telstra, Optus, and TPG, probes inside each network show asymmetries that no single commercial provider can see.

What volunteer probes actually measure

Volunteer probes are simple by design. They run network tests against target domains and report outcomes. Tests include DNS resolution, HTTP HEAD requests, and TLS handshakes. A probe behaves like a normal client on the network, which is the entire point.

Because probes are passive from the network operator's perspective, they evade the cat-and-mouse game that commercial VPN endpoints trigger. An ISP that throttles VPN protocols may still serve normal traffic to a probe, revealing the true baseline. Researchers comparing results from Adelaide, Hobart, and Darwin can spot regional differences in DNS hijacking.

Precision in measurement requires the same rigour found in regulated fields. The methodical standards seen in guides to obtaining a Part 107 waiver for night operations echo the documentation discipline we apply to our probe network. The software is open source, and configuration files are public. This auditability is non-negotiable. A black-box system cannot replace a transparent one when findings may inform public policy.

Geographic coverage and the Australian context

Australia's internet infrastructure is concentrated on the eastern seaboard, with Sydney, Melbourne, and Brisbane hosting most peering exchanges. Perth and Adelaide sit on the edge of the domestic backbone, often routing traffic through Singapore for international destinations. This topology creates blind spots for commercial VPN providers.

Volunteer probes fill those gaps. A probe running on a home connection in Fremantle or Cairns sees the same routing table and DNS resolvers that a resident's laptop sees. When the probe records a DNS failure or TCP reset, it reflects a real condition on that link. The aggregated dataset shows how filtering behaves across the National Broadband Network and the Optus cable footprint.

Commercial VPN nodes in Australia are frequently overwhelmed during peak evening hours, when households stream sport. Probes sit idle most of the time and run tests on a schedule, producing a less noisy dataset.

Our volunteer community includes people who balance technical contributions with personal wellbeing. We point newcomers toward resources that support their broader lives, including a trauma-informed yoga practice for those who process difficult online content as part of their work.

Methodology transparency

Commercial VPN providers treat measurement methodology as a competitive advantage. They rarely disclose how they sample targets, rotate endpoints, or filter false positives. The result is a dataset that is internally consistent but externally opaque.

Volunteer probes follow the opposite philosophy. The test list is public, source code is public, and raw output is publishable in near-real-time. Researchers can replay tests, audit the code, and reproduce results.

This transparency helps us defend our findings against criticism. When the Australian Communications and Media Authority or a major carrier asks how we reached a conclusion about a blocked domain, we can point to specific probe logs, IP addresses, and timestamps. That level of evidentiary support is difficult to achieve with commercial telemetry.

Limitations of commercial VPN datasets

Commercial VPN services do have legitimate uses. A user choosing between providers wants to know latency, uptime, and streaming compatibility. Those metrics require running real traffic through the VPN tunnel.

The problem arises when commercial datasets are repurposed for censorship research. A VPN endpoint in a Sydney data centre may not see the same DNS resolvers as a household in Parramatta. A commercial dataset overrepresents networks where the provider has nodes, biasing coverage toward major metropolitan areas.

There is also a structural conflict of interest. A commercial VPN provider earns revenue by promising its service bypasses restrictions. Publishing data showing certain restrictions are uncircumvented would undermine that promise. Volunteer probes have no such incentive.

Specific shortcomings include:

• VPN nodes concentrated in capital cities, missing Newcastle, Geelong, or the Gold Coast
• Aggregation levels that hide individual probe paths
• Test schedules optimised for performance, not censorship detection
• Incentives to report fewer restrictions than actually exist

Comparing the two approaches

Aspect	Volunteer probes	Commercial VPN data
Measurement point	End-user network paths	Provider data centre endpoints
Transparency	Open source, public logs	Proprietary, aggregated
Geographic spread in Australia	Distributed across suburbs and regions	Concentrated in capital city data centres
Conflict of interest	None, volunteer-run	Commercial incentives
Resistance to detection	Behaves like normal traffic	Often flagged by deep packet inspection
Cost to maintain	Community donations	User subscriptions

The table summarises structural differences. Each row reflects a design choice that affects suitability for censorship measurement.

Reading the statistical charts

The censorship dashboard displays a range of statistical outputs. A common mistake is to treat a single probe failure as evidence of blocking. A careful approach considers the confidence interval, the number of probes that tested the target, and the consistency of the failure mode.

For researchers and journalists, we suggest the how to interpret the statistical charts on the censorship dashboard guide, which covers time-series plots and the difference between transient failures and persistent filtering.

Key indicators to watch include:

• The percentage of probes reporting a given result across a time window
• The variance between probes in the same autonomous system number
• The presence of DNS resolver inconsistencies
• The duration and consistency of any anomaly

Resilience and community trust

A volunteer probe network is only as strong as the community behind it. We rely on volunteers who run probes in homes, offices, and on portable devices. The diversity of connections, from NBN fibre to fixed wireless in regional Tasmania, gives the dataset a texture that no commercial fleet can replicate.

The project is supported by donations and volunteer labour, with data provided for indicative research purposes. We make methodology, code, and raw outputs available through neumon.org so that anyone can verify or challenge our findings. That openness is the foundation of trust in an area where accusations of bias are common.

Volunteers are taking on a small civic responsibility. They participate in a distributed measurement of public infrastructure, contributing to a shared picture of how the Australian internet actually behaves.

Volunteer probes will never replace commercial VPN services for privacy or streaming. For censorship measurement, the evidence points toward distributed, transparent, community-run infrastructure. The practical takeaway is that distributed measurement beats aggregated commercial data for understanding how the Australian internet behaves, and the quality of that understanding improves with every new probe added to the network.