Use GitHub's private advisory form. It opens a report only maintainers can see, and it is the right channel for anything below.

If that is not available to you, write to the Telegram contact in the README and say only that you have a security report — do not put the details in the first message.

Please do not open a public issue for a vulnerability. Everything else belongs in a public issue, including bugs that merely look alarming; if you are unsure which one you have, use the private form and we will move it out if it is fine in the open.

There is no bounty. There is no company behind this and no budget to run one, and saying so plainly is better than a page that implies otherwise.

This is an anti-detect browser, so "security" means two different things and both matter.

The usual kind. Anything that lets code that should not have access get it: a page reaching the agent's IPC channel, the local HTTP API answering without its token, a profile bundle decryptable without the key, a server endpoint returning another organisation's rows, a proxy password recoverable from disk, an escape from Chromium's sandbox that our patches introduced.

The kind specific to this project: a fingerprint that gives the profile away. If you can distinguish a Fury profile from the browser it claims to be, that is a real defect and we want it. Concretely:

• a value the persona sets that some context reports differently — a Worker, a ServiceWorker, an iframe, an AudioWorklet. Cross-context disagreement is the single most valuable class of report, because it is what real detectors look for and it is invisible from the main frame.
• a property that reveals automation, a patched engine, or Fury itself.
• a real value leaking through a spoofed one — a timezone, a locale, a font list, a device, an address.

A working page that demonstrates it is worth more than a description, and tools/detect-suite/ is the harness we use ourselves.

Not a vulnerability: that a site detects a proxy, a datacentre IP, or an account's behaviour. Fury spoofs a browser, and never claimed to spoof a network or a person. The README's "Not done yet" table lists what is known to be missing, and something already on it is not a new finding — though a measurement that makes it worse than the table says certainly is.

You will get a reply. If the report is a fingerprint difference, the reply will usually include whether we could reproduce it, because a claim about a browser that nobody has run is not worth much in either direction.

Fixes land as a patch in core/patches/ with a verify script beside it in core/verify/, so the fix and the proof that it works arrive together. That is the house rule for everything here and it applies to security fixes in particular: a patch that applies and compiles has been wrong about the browser before.

Credit in the commit and in the release notes, under whatever name you want, including none.

None yet — there has been no release. Until there is, the supported version is main, and a report against anything else cannot be acted on.

Worth stating, because an anti-detect browser attracts the assumption that it does more than it does.

It hides how your browser looks. It does not hide who you are: an account you log into identifies you, a payment identifies you, and a pattern of behaviour identifies you across every profile that shares it. It is not Tor and is not a substitute for it. It does not make anything legal that was not.

ACCEPTABLE_USE.md says what we ask of you. It is a request, not a technical control, and it is deliberately not enforced in code — a browser that decides what you may visit is a browser with an opinion, and one that reports the decision is worse than the problem.