Privacy Signals Check - Permissions and Automation
docReview site permissions, automation indicators, and privacy-related browser signals to quickly identify tracking exposure and risk control issues.
Privacy checkers usually make one claim and imply a bigger one. They measure what a website can learn about you from a single request, then leave you to guess the rest. The honest version of the claim is narrower: some of what a site sees is fixed by the network, some is announced by your browser, and some is probed by code. A privacy page that shows all three is more useful than one that shows a number and stops.
Three channels a site can read
| Channel | What it carries | Can you change it? |
|---|---|---|
| Network layer | Source address, ASN, country, TLS handshake | Only by changing the exit |
| Announced headers | User-Agent, Accept-Language, Referer | Partly, in browser settings |
| Client-side probing | Screen size, timezone, fonts, WebGL, canvas hash | Rarely without breaking sites |
The first channel is the hardest to reason about, because the address a site sees depends on the path between you and it, not on the address your machine reports locally. The second is often outdated or generic. The third is the one people underestimate: it is cheap, it is precise within a browser build, and it survives most network changes.
What "privacy" actually means in this context
There is no single score. A page can measure exposure without saying whether exposure is bad:
- Identifiability. How unique your combined signals are among a sample. A unique combination is a tracking handle even without a name attached.
- Linkability. Whether the same signals recur across sessions and sites. This is what allows a profile to accumulate.
- Persistence. What survives a restart: cookies, local storage, IndexedDB, and settings. Clearing one does not clear others.
- Network exposure. What the exit reveals about your location, your provider, and whether the traffic is tunnelled.
A checker that only reports an IP tells you about the fourth and nothing about the first three. That is a partial picture, and calling it a privacy score is a stretch.
Storage is the part people forget
Cookies are the familiar mechanism, and browsers now constrain them heavily. The less visible stores are not constrained the same way:
- Local storage persists until explicitly cleared, per origin. It is the default home for app state and identifiers.
- IndexedDB holds structured data and can be large. It survives cookie clearing.
- Cache storage can retain responses and is rarely inspected by users.
- Browser-level identifiers such as installed fonts, extensions and settings are readable in aggregate through probing, without any store at all.
Practical implication: "clear cookies" is one step, not a reset. A consistent session across sites after clearing cookies is usually one of the other stores, or a stable fingerprint, not a cookie that survived.
What a privacy check can and cannot prove
Can prove with reasonable confidence:
- Which address and ASN a request originated from, and which country the registry associates with that block.
- Whether headers announce an inconsistent locale or a generic user agent.
- Whether the browser is exposing a set of fingerprintable properties in the clear.
Cannot prove:
- That a particular site is tracking you. That requires observing the site's behaviour, not the client's properties.
- That a signal is unique. Uniqueness within a small sample overstates it.
- That changing one property improves privacy overall. Changing an isolated field often makes the combination more distinctive, not less.
Reasonable practice
- Separate the questions. "What does this network reveal?" and "what does this browser reveal?" have different fixes.
- Prefer controls that do not break the web. Blocking third-party cookies and limiting cross-site storage are supported by browsers; randomising headers is not.
- Treat a single number as a prompt, not a verdict. Use it to decide what to look at, not what to conclude.
- Re-check after a change. A new tunnel changes the network picture and usually nothing else; a new browser build changes the fingerprint and leaves the network alone.
Common misreads
- "The page says my privacy is low, so I am being tracked." It says your signals are distinctive. Distinctiveness is a precondition, not evidence of tracking.
- "I use a VPN, so the privacy score should be high." A tunnel changes the network channel. The browser channels are untouched.
- "Clearing cookies resets everything." Local storage and IndexedDB usually survive, and a stable fingerprint needs no storage.
- "A fingerprint is a person." It is a combination of properties shared by a population. Precision is not identity.
FAQ
Is a privacy score meaningful on its own?
Only as a summary of measured exposure. It is not a risk assessment, and it cannot see what a specific site does with the data.
Does incognito mode change the result?
It isolates cookies and storage for the session. Network signals and fingerprintable properties are unchanged, so the exposure picture is largely the same.
Can I be identified without cookies?
Within a browser build and a network, a distinctive combination can function as a handle. It is less reliable than a cookie but it does not require consent.
Should I randomise my headers?
Usually not. Malformed or implausible headers are themselves a signal, and modern browsers expose less through headers than they used to.