multi-account-isolation

Profile Isolation - verifying it, not assuming it

A profile that looks isolated usually is not. The failures are boring and mechanical: a timezone that does not match the exit IP, a WebRTC candidate carrying the real address, a canvas hash that changes on every read, two profiles that ended up on the same persona. This skill is the check list for catching those before they matter.

Authorized use only. This is for identities you own or are authorized to operate: your own accounts, your own test fixtures, your own QA fleet, and your own anti-fraud stack. It is not for accessing systems without authorization, for accounts that are not yours, or for creating fake accounts or engagement. Comply with the terms of the sites you automate and with applicable law - see Acceptable use.

What this does not claim. Passing every check below means the browser layer is internally consistent. It does not mean a given site will treat two profiles as unrelated: things entirely outside the browser - a shared payment instrument, a shared contact detail, identical activity patterns - are not something any browser setting reaches. Treat a clean result as "the technical layer is not the problem", not as a guarantee.

For the SDK that creates and launches these profiles, see the anti-detect-browser skill.

The configuration invariant

One identity gets one of everything. Any cell shared between two identities is a defect to find:

identity  →  profile  →  persona  →  proxy  →  timezone
   1      :     1     :     1     :    1    :     1

Profiles are unlimited and free on every antibrow plan, so there is never a reason to reuse one. "Log out and log back in as the other identity" inside one profile defeats the entire setup - the cookie jar and localStorage are the point.

Setup under test

import { AntiDetectBrowser } from 'anti-detect-browser'

Python, same on-disk profile format:

The checks

Run each profile through its own proxy, and assert rather than eyeball.

Checks 1, 3 and 7 are the ones worth wiring into CI: they are cheap, deterministic, and they catch the defects that actually recur.

Reading a failure

Work down in this order, cheapest first - a fingerprint is almost never the actual cause:

1. Profile name reused? list_profiles, or compare browser.profile_dir per identity. Two identities in one directory explains everything else. Directories are named after the profile's id, not its name, so match on profile.json inside rather than on the folder name.
2. Same address twice? Confirm each public_ip is distinct.
3. Clock disagrees with the address? Print browser.timezone and browser.public_ip together.
4. Persona regenerated? If the canvas hash moved between launches, the profile is not frozen - check whether profile_dir or the cache directory changed under it.
5. Only then the fingerprint itself, verified with the suites above rather than assumed.

What the runtime touches, and how to check it

Any tool that drives logged-in sessions receives cookies and proxy credentials, so it is fair to ask what it does with them. For antibrow:

The kernel is a closed-source Chromium build - that is the tradeoff for the spoofing living in C++ rather than in an injectable script - so verify behaviour rather than take it on faith:

Point it at a proxy whose logs you can read, or at a local MITM proxy, and watch what leaves the machine during a launch. Pin the SDK version and check the published hash (npm view anti-detect-browser@2.8.0 dist.integrity) so the code you audited is the code that runs. If a deployment must not phone home at all, this is the wrong tool: license verification is compiled into the kernel and there is no offline mode.

What isolation cannot cover

Worth stating plainly, because a clean check list invites the wrong conclusion:

• Anything outside the browser. A shared payment instrument, a shared contact detail, a shared payout destination - no browser setting touches these, and they are the strongest correlators that exist.
• Activity patterns. Identical timing, identical content, identical interaction targets. Not a technical property.
• Identity verification. A document check is not a fingerprint problem.
• A platform's own decision. Nothing here changes how a site chooses to treat an account.

If every check passes and something still looks wrong, the cause is in this list, not in the browser layer.

Acceptable use

Intended: verifying isolation between identities you own; running client accounts with the account holder's authorization; building QA fixtures that emulate distinct devices; testing your own anti-fraud and correlation logic; auditing what a browser runtime does with your credentials.

Out of scope, and not supported: accessing any system without authorization; logging into accounts that are not yours; credential stuffing or account takeover; creating fake accounts, reviews or engagement; circumventing an authentication, payment or authorization control; scraping personal data in violation of applicable law; working around a platform's enforcement decision.

Complying with the terms of the platforms being used, and with applicable law, is the operator's responsibility. Report abuse or a security issue via the contact at https://antibrow.com.

Related Skills

• anti-detect-browser - the SDK, profiles, personas, proxies and REST API that create the setup being verified here
• browser-mcp-agent - MCP server mode, for letting an AI agent drive a single profile itself