Home › Protection
How an address leaks, and what actually helps
A long read with no product at the end of each paragraph. Most of it applies whether or not you ever buy anything here.
Contents
1. The six ordinary paths into a database
Addresses do not escape by magic. There are a small number of routes and they are worth naming, because each one has a different counter.
Checkout fields. Buying anything means handing over an address, and the merchant's terms usually permit "partners". The field is mandatory even when the receipt could go by other means.
Loyalty and warranty cards. A warranty registration exists to identify a purchase to a person. That is exactly the shape of a marketing record.
Competitions. Entry is free because the entrant list is the prize. Terms almost always allow transfer.
Public posting. A forum profile, a mailing-list archive, a business card in a PDF. Once it is indexed, harvesting is trivial and permanent.
Other people's address books. Someone forwards a group message with everyone visible, or installs an app that asks for contact access. Your address enters a database you never agreed to.
Breaches. The one route that requires no action from you at all.
Note the pattern: five of the six are collection with consent, granted years ago, in a checkbox nobody read. Consent is not the problem. Scope is.
2. Resale and the master record
When a list is sold, the buyer does not receive your unsubscribe state — they receive a row: address, maybe a name, maybe a rough segment, a date. The original subscription's rules do not travel with it. Unsubscribing from the copy has no effect on the original, which is why the mail comes back after a year under a different brand on a domain registered last month.
This is the structural reason a single address fails. It is one credential reused across two hundred relationships, and the weakest of those two hundred decides the fate of all of them. One careless reseller at number ninety-four exposes the address to a hundred more, and the other two hundred never find out.
Compartmentalising fixes the shape of the failure. Give each relationship its own address and a resale becomes visible and local: the junk arrives addressed to the specific alias, so the seller is identifiable within a day, and that alias can be burned without touching the other 199.
3. Tracking pixels, in detail
A tracking pixel is an image tag one by one pixel in size, with a unique URL per recipient. When your client fetches it, the sender learns four things: that the message was opened, when, roughly where from (by IP), and what client you use. Fetch it twice and they learn you reread it.
This is not forbidden and not usually malicious. It is how senders decide which subject lines work. The cost is a behavioural record you did not consciously create, and it is often shared with an analytics vendor who has the same record for fifty other senders.
Defences, in ascending order of effectiveness:
- Block remote images. Free, in every client. Breaks layout and hides product shots.
- A client that proxies images. The proxy fetches, so the sender sees the proxy. Better, but tied to one app.
- A filtering layer that proxies before delivery. Works regardless of client. Loads images on request. This is what the service does.
- Plain-text reading. Nothing to fetch at all. Thorough and joyless.
Worth knowing: "open rate" is inflated almost everywhere by automated prefetching, so a sender's numbers are wrong in the same direction for everyone. Your own record is still being built.
4. Breach dumps and credential stuffing
A dump is a database stolen and published. Addresses are the cheap part of it — they are worth very little alone. The valuable part of a dump is the pairing: this address, this password hash, this site.
Credential stuffing takes those pairs and replays them against other sites, on the correct assumption that people reuse. A breach at a bicycle forum in 2019 is a working key to an account that used the same password in 2026. The chain is long and nobody remembers the first link.
Two consequences follow. First, the address in a dump is permanent — you cannot un-publish a file already on a hundred mirrors. Second, the useful response to a leak is not a new address, it is a password change on every account that shares the exposed one, plus a recall of the alias if the leaked site is still mailing you.
Monitoring matters here mainly for speed. A leak you learn about in March is a repair; the same leak learned in November is sometimes already an incident.
5. What measurably reduces risk
- One alias per service. Contains resale, names the seller, makes recall surgical.
- Recall on first abuse. A bounced alias gets pruned. Senders do respect a hard bounce more than an unsubscribe link.
- Proxied images. Removes the open record without breaking the mail.
- Unique passwords in a manager. Turns a dump from an incident into an inconvenience.
- Second factor on the mailbox itself and on the accounts that can reset it. The mailbox is the master key; everything else is secondary.
- Reducing the collection surface. Optional fields left empty, loyalty cards refused, marketing checkboxes unticked.
- Faster response to alerts. The single biggest variable in how much a leak costs you.
6. What looks like protection but is not
"Private mode" in a browser. Stops local history, does nothing about a pixel fetched by the mail client.
Unsubscribing from every list. Confirms the address is live and read. Useful, but it is a cleaner, not a shield — and it does not touch the master record.
Deleting an app. The account survives, and usually the address with it.
Spam filters alone. They move the symptom out of sight. The address is still being sold and still being counted, and filter rules drift.
A second address used "for junk". Better than nothing, but it concentrates every signup into one new record that becomes as poisoned as the first, usually within a year.
Antivirus. Genuinely necessary, genuinely unrelated. It scans files, not address lists. This is why the footer of every page here says the same thing.
7. Passwords and two-factor, briefly
Three rules carry most of the weight. Use a manager so no two accounts share a password. Give the mailbox itself the strongest unique password you own, since it can reset everything else. Put an authenticator or hardware key on it, then on the accounts that can recover it, and prefer app-based codes to SMS where both exist — SIM swapping is real and boring.
Store recovery codes somewhere offline. A second factor you cannot recover is a lockout waiting for a bad week.
8. A checklist for this week
- Turn off remote image loading in your mail client, or get proxying in place.
- Pick the ten accounts that matter most and give each a unique password.
- Add a second factor to the mailbox and to the two accounts that can reset it.
- Write down recovery codes and put them somewhere that is not the mailbox.
- Unsubscribe from the five lists you never read — then recall their aliases.
- Check your address against a public breach index and change any reused password found.
- Empty optional address fields on the next three checkouts.
- Before the next sign-up, request an alias rather than typing the real address.
Assistance service, not an antivirus. Where the guide gives general advice it is offered for information; the state of a particular mailbox is established during the audit, not from this page.