groundy
Infrastructure & Runtime

After HN Email Scraping: Per-Site Aliases vs Plus-Addressing for Developers

Developers should use unique per-site aliases for new public surfaces to enable attribution and revocation, while treating plus-addressing as mere labeling that exposes the re

Published 6 references
Two imperfect glazed ceramic mailbox forms on ivory: a green one with a detached spout beside an open socket, and a black one with an exposed spout and loose yellow collar. Hard shadows emphasize their handmade texture.
On this page9 sections

If your email address is already public on Hacker News, in a commit history, or on any other long-lived public surface, treat it as burned: rotation can contain future exposure, but it cannot retrieve addresses already harvested or undo associations already made. For new hiring posts, project announcements, and site registrations, I would use a unique per-site alias rather than plus-addressing. A unique alias can identify which surface received an address and can be retired independently. A +tag remains attached to a recoverable base address, so it is useful for mailbox organization but weak as a privacy boundary.

That recommendation does not depend on proving a large scraping campaign. The immediate trigger is an anecdotal Tell HN thread observed on November 16, 2026, not a measurement of prevalence. The phishing research below is also author-reported and not independently replicated. Together, these sources support a containment decision; they do not establish how often HN scraping occurs, confirm the suspected data-broker pipeline, or compare alias providers head-to-head.

What the HN reports establish, and what they do not

The strongest report in the thread is also the simplest. A commenter said automated, template-based messages had arrived roughly once a day for many weeks at the unique address placed in a “Who Wants To Be Hired” comment. Because that address was unique to the post, subsequent mail could be attributed to that exposure surface without knowing who sent it or how the address was obtained.

Other reports were less controlled. One user reported hundreds of junk messages in the weeks after posting an open-source project, including monetization offers and fake hosting-provider scams. Another described an email sent to an address associated with a seven-year-old throwaway HN account containing a “scary amount of context.” That recipient wrote:

I’m guessing the e-mail address appears in a separate database sold by a data broker, and they were able to use that to contextualize the e-mail.

The guess matters. A contextualized message demonstrates that the sender had or constructed relevant context; it does not demonstrate that a broker sold a particular database, nor that HN was the original source of every association. Likewise, junk mail after an open-source announcement is consistent with scraping but does not exclude other collection routes.

For developers, the practical inventory is therefore broader than the evidence. Hiring-thread comments and public project announcements appear directly in the reports. Commit histories and package manifests are also worth auditing because they can carry durable contact addresses, but this thread does not measure scraping from those surfaces. The defensible assumption is not “HN is the sole leak.” It is “any address published on a long-lived public surface may persist beyond my control.”

Why better filtering is only part of the answer

Context makes phishing harder to assess by content alone. A generic template can exhibit obvious bulk-mail patterns; a message that accurately references a recipient’s project, work history, or public identity has fewer generic signals to distinguish it from a genuine approach.

The SpearBot preprint illustrates the generation side of that problem. The preprint’s authors report a framework combining LLM jailbreak procedures with critique-based optimization. They curated 1,000 spear-phishing emails covering 10 phishing strategies and reported that human evaluators found the generated messages significantly readable and deceptive. That is evidence about generation and human evaluation, not a measured bypass rate against any particular production spam filter.

Detection research also argues against treating a specialized detector as a complete boundary. The PiMRef preprint describes an approach based on sender-identity knowledge. The preprint’s authors found that messages could evade the method when sender identity was highly ambiguous or absent, such as abbreviating Kaiser Permanente as “KP,” although that evasion sharply reduced plausibility. This is a constraint on attackers, not a guarantee that evasive messages will be caught.

The evaluation scope narrows the claim further. PiMRef’s field study involved three volunteers, and its university spam feed contained mail already flagged by RSpamd. Its measured recall therefore concerned that pre-filtered set rather than the full incoming mail stream. Its domain-level knowledge base, shared across users, also cannot represent every address-level exception where a particular sender is legitimately authorized to represent an identity.

None of this makes filtering useless. It changes its job. Filtering remains worthwhile for reducing volume, while address design limits how one exposed contact point can be reused and helps the recipient investigate what remains. Accurate personal context should not itself authenticate a sender, especially on an address with years of public history.

There is also a counterweight to focusing exclusively on highly tailored attacks. The paper Targeted Attacks: Redefining Spear Phishing and Business Email Compromise cites Abnormal Security figures from the second half of 2021: approximately 76.7% of attacks against organizations were ordinary phishing, while 3.81% were targeted attacks. Those figures predate current LLM phishing tools and are not current rates. They nevertheless caution against dismantling ordinary spam defenses in response to a few alarming, context-rich examples.

Plus-addressing is labeling, not containment

Plus-addressing appends a tag to an existing mailbox name. In a hypothetical example, dev+hn@example.com identifies dev@example.com once the +hn portion is removed. Anyone handling the tagged form can derive that base string; no breach of the provider is required.

SimpleLogin’s documentation makes exactly this criticism: removing the plus portion reveals the real address, a tagged address in a breach can expose the base address, and removing tags lets advertisers link addresses together. SimpleLogin sells an aliasing service, so its broader privacy framing is commercially interested rather than an independent comparison. The string-removal mechanism itself is straightforward, but the documentation does not measure how often advertisers or scrapers perform that linking in practice.

The distinction for containment is what happens after the first exposure:

  • A tag can help the recipient organize or label incoming mail.
  • It does not keep the base mailbox address out of the exposed string.
  • Once later mail is sent directly to the base address, the original tag cannot reliably identify which surface enabled that message.
  • Filtering or rejecting one tagged form does not retract the base address already derived from it.

This does not mean developers must stop using plus-addressing. It means they should budget it as convenience tagging on an address they are already willing to expose. It is the wrong layer when the requirement is to keep a stable primary address private or to revoke one site’s contact route without affecting the underlying mailbox.

What per-site aliases actually buy

A per-site design assigns a distinct address to each public post, service, or other exposure surface. One HN commenter described a plan for a dedicated domain containing throwaway, single-use addresses per site, accompanied by a publicly accessible page recording which addresses had been compromised or received spam. That was a to-do list, not a reported deployment or evaluation, but it captures the two useful properties.

First is attribution under controlled reuse. If an address is used for only one hiring comment, later mail to that address identifies a connection to that comment’s exposure surface. It does not identify the scraper, prove whether collection was manual or automated, or reveal subsequent sales and sharing. Attribution becomes weaker as soon as the same alias is reused across sites.

Second is independent revocation, provided the chosen alias system allows addresses to be disabled separately. Retiring one alias can cut off that contact route without changing every other site’s address or abandoning the destination mailbox. This controls future delivery through the alias; it does not erase messages already received or remove the address from copied lists.

Opaque local parts also avoid publishing the base address in the literal way plus-addressing does. They should not be mistaken for guaranteed anonymity. If every alias shares one dedicated domain, that domain remains common information available to anyone comparing the addresses. None of the cited studies tests cross-site linkage against real alias deployments.

A compromise log can be useful without becoming a claim of complete detection. Recording the surface, the alias used, and when unwanted mail began preserves evidence that would otherwise disappear into a crowded inbox. The thread supplies the idea of making such a page public, but nothing cited evaluates that choice. The operational benefit comes from maintaining the record, not from publicity itself.

An alias domain is observable too

Running unique addresses on a domain does not make the domain invisible to list-verification tooling. The open-source listwash repository documents SMTP mailbox probing that issues RCPT TO and resets without sending DATA, so no message is delivered. It reports servers that block probes, greylist, or answer ambiguously as unknown rather than guessing, and its description advertises catch-all detection.

There is no evidence here that the HN senders used listwash or any particular verifier. The tool instead demonstrates an available technique: systems can probe mailbox behavior and attempt to fingerprint catch-all configurations without sending a visible email.

Consequently, a catch-all or dedicated domain should not be treated as a secrecy mechanism. Its defensible purpose is to make exposed addresses attributable and individually disposable even when an observer can identify the domain or probe its behavior. An “unknown” verification response also should not be read as proof that an address is undeliverable or that probing has been defeated; listwash explicitly uses that classification for ambiguous server behavior.

The larger implementation choice remains under-evidenced. Neither the thread nor the cited papers establish self-hosted deliverability, account-recovery behavior, migration costs, provider lock-in, or current provider-specific alias limits. Before putting important accounts behind a dedicated domain or an aliasing provider, verify current documentation and test the recovery path. The questions that can change the choice are:

  • What happens to essential mail if the domain or provider is unavailable?
  • Can addresses and their routing be moved without silently losing access?
  • Which accounts would become unrecoverable if their only registered contact address stopped working?
  • Can one alias be suspended without disturbing unrelated addresses?

Those are unresolved decision inputs, not reasons to retain plus-addressing as a privacy substitute. Until they have satisfactory answers, I would avoid making a new alias domain the sole recovery route for critical accounts.

Match the layer to the exposure

Exposure situationAppropriate layerWhat it can contain or revealWhat it cannot undo or establish
An address already present in old posts, commits, or manifestsContinued scrutiny and filtering on the existing mailboxReduces reliance on contextual trust; preserves monitoring of mail still arriving thereCannot retract distributed copies, broker-held associations, or the address’s public history
A new public hiring comment or project announcementOne unique alias for that surfaceAttributes later mail to that surface; allows that contact route to be retired independentlyDoes not identify the collector, prove scraping prevalence, or prevent template spam from being generated
A new site registration where the primary mailbox should remain undisclosedAn opaque per-site aliasAvoids exposing the base address through a removable tag and limits reuse of one contact routeDoes not guarantee unlinkability, erase prior exposure, or establish that the implementation is recoverable
Low-risk labeling on an address already acceptable to exposePlus-addressingConvenient tagging and mailbox organizationDoes not conceal the base address or provide durable attribution after senders strip the tag
A choice between self-hosting an alias domain and using a providerCurrent documentation checks and recovery testing before migrationResolves implementation-specific operational questionsThe HN thread and cited studies do not compare deliverability, recovery, portability, or lock-in

“Burned” should not be read as “delete the mailbox immediately.” Established contacts and old account-recovery paths may still depend on it. It means withdrawing the assumption that the address is private and avoiding new uses that require privacy.

For every new surface, the stronger default is a unique per-site alias, implemented either on a dedicated domain or through an aliasing provider whose current operating and recovery terms have been checked. Keep plus-addressing for organization, not identity separation. Continue filtering and suspicious-message review on old addresses, because no alias system can unsell an existing association.

The evidence supports that architecture more strongly than it supports any particular product. The HN reports demonstrate why unique addresses are useful, SimpleLogin documents the reversible plus-tag mechanism, and the phishing studies bound a detection-only strategy. None measures how much aliasing reduces phishing in practice or independently tests the two approaches against scraping. The decision is therefore containment based on mechanics and bounded inference, not a promise that a new domain will make a public identity private again.

Frequently Asked Questions

What is the main difference between per-site aliases and plus-addressing for privacy?

A unique alias can identify which surface received an address and can be retired independently. A +tag remains attached to a recoverable base address, so it is useful for mailbox organization but weak as a privacy boundary.

Does plus-addressing hide the base email address?

Plus-addressing appends a tag to an existing mailbox name. In a hypothetical example, dev+hn@example.com identifies dev@example.com once the +hn portion is removed. Anyone handling the tagged form can derive that base string; no breach of the provider is required.

References

Follow the links in the article for context. The supporting material is collected here for further reading.

  1. anecdotal Tell HN thread observed on November 16, 2026news.ycombinator.comAccessed
  2. SpearBot preprintarxiv.orgAccessed
  3. PiMRef preprintarxiv.orgAccessed
  4. SimpleLogin’s documentationsimplelogin.ioAccessed
  5. listwash repositorygithub.comAccessed

Join the discussion

Share a useful perspective or ask a question about this article.

Discussion guidelinesComments privacy