Back to blog

How to Assess WebRTC IP Leaks: Candidates, Proxy Egress, and Retesting

An address on a WebRTC test page does not automatically mean your real public IP has leaked. Learn to read host, srflx, relay, and mDNS results, record ordinary web egress and browser/network conditions, use repeatable comparisons to identify unexpected exposure, and verify calls after privacy-policy changes.

When a WebRTC test shows an address different from ordinary web egress, start with its type and the test conditions. Private addresses, masked local hostnames, and unexpected public egress mean different things. A red warning alone cannot settle the issue.

A reliable check asks which egress ordinary pages use, what candidate information WebRTC exposes, whether that information exceeds your expectations, and whether changing settings actually changes the results.

Why does WebRTC produce multiple candidates?

Real-time audio, video, and data connections need possible communication routes. ICE candidates describe addresses and connection methods that can be tried; the candidate list is not the communication path ultimately selected.

MDN’s candidate type documentation distinguishes host, srflx, prflx, and relay. These types organize the interpretation below; none is automatically a safety or danger label.

Result typeBasic meaningWhat to check next
hostA candidate associated with the local hostPrivate address, public address, or masked name
srflxA server-reflexive address discovered through a serverWhether unexpected network egress appears
prflxA peer-reflexive candidate found during connectivity checksIts appearance conditions and the path actually used
relayA candidate supplied by a relay serverWhether the relay is expected and calls work

Even unselected candidates can expose information worth inspecting. Conversely, a large candidate count does not by itself prove accounts have been linked or a platform has identified you.

What do private addresses and mDNS names mean?

Common private addresses beginning with 192.168 cannot be directly compared as internet egress addresses. They may reveal LAN information, but that differs from revealing an ISP’s public egress. State the specific concern when investigating.

A result ending in .local may be an mDNS-related name rather than a directly readable original LAN IP. Microsoft’s mDNS address-masking documentation explains masking and JavaScript visibility. Its enterprise configuration is not a default recommendation for ordinary privacy testing.

Do not copy enterprise instructions to remove masking just to get “more complete” test results. Nor does .local mean all network information is hidden: inspect reflexive, relay, and other candidates and the actual browser settings.

Establish a comparable network baseline

Record browser and version, operating system, proxy or VPN scope, split routing, and ordinary web IPv4 and IPv6 egress. Do not expose full addresses in public tickets or screenshots; keep necessary details only in your own controlled records.

Run the WebRTC check in the same browser environment with the same network and permission conditions, and wait for candidate gathering to finish. Record address categories, protocols, and timestamps rather than just capturing “pass” or “fail.”

Whether a proxy covers all relevant browser connections must be verified with actual results. Correct ordinary web egress establishes only that request’s path; it does not prove that every real-time connection follows it.

How to make a meaningful comparison

Save current results and change only one factor at a time, such as proxy configuration, network routing, or browser privacy policy. Restart the relevant working environment or connection, observe again under the same test conditions, and verify the change is repeatable.

If WebRTC still exposes a confirmed unexpected public address under conditions where you intended to hide it, investigate the associated interface, IPv6 routing, and proxy coverage. Do not judge solely by the city in an address database, which may be inaccurate.

If you need a no-proxy comparison, choose whether to run it according to privacy needs only in your own trusted test environment. No business account is needed, and testing does not justify visiting untrusted pages. If direct networking is unacceptable, compare with routing information from the network administrator.

Before changing policy, check who controls it

Chrome’s privacy API documentation provides WebRTC IP handling policies that balance routing, address exposure, and media performance. It also explains that enterprise policy or other extensions may control settings; selecting an option in the interface does not necessarily mean it takes effect.

Check whether organizational policy manages the browser and whether extensions compete for the same setting, then follow that browser’s official instructions. Do not install multiple “leak fix” extensions simultaneously or assume experimental flags in old tutorials remain valid forever.

For multiple work-specific browser environments, see environment and proxy management on the PurpleMark website. Recording each environment’s proxy and owner ties retest results to a specific configuration and reduces confusion from mixing environments.

Empty results, disabled functionality, and working calls are different

No candidates may result from feature restrictions, network blocks, page errors, or insufficient test conditions. It does not prove full device anonymity. Check that the page executes correctly and repeat under the same conditions instead of collecting only favorable screenshots.

Disabling related functionality completely may affect meetings, customer support, livestreaming, or collaboration tools. After changes, test audio, video, and data connections in your own test scenarios. If relying on a relay, also check that latency and stability meet real work requirements.

Do not optimize only for one test page’s score. Define a verifiable goal, such as “no unexpected public egress exposure while authorized meetings work,” and decide whether to keep the configuration from the results.

When should you retest?

Browser upgrades, proxy replacement, VPN split-routing changes, network switches, and extension updates may all require retesting. Record the reason for changes and results to distinguish configuration regressions from changes to the test page itself.

A retest report should retain at least environment ID, version, network scope, candidate categories, expected versus actual differences, and call-function results. You do not need to publish complete ICE strings, personal addresses, or credentials when seeking help.

Assessing WebRTC leaks depends on explaining specific addresses and repeatable conditions. Identify the exposure first, choose suitable routing and privacy policies, and finally verify both privacy goals and actual communication. That completes the investigation.