Uncategorized

Stop Hard-Rejecting Every Disposable List Match

Most sign-up abuse defenses fail in the same spot: a single blocklist hit triggers a permanent rejection. That works until a real customer shows up on a domain that some list marked disposable months ago, or until a shared hosting provider gets flagged because one tenant ran a temp-mail service. The fix is not to stop checking. The fix is to stop treating every match as equal. Log which source matched, score the risk, and give the user a path back in through email verification.

Rejecting on a single source is the first mistake

Every disposable email domains list has a different curation model, refresh cadence, and false-positive rate. Treating them as interchangeable is how you end up blocking a corporate domain that happens to share infrastructure with a burner provider. The disposable-email-domains project is a community blocklist plus an allowlist, which means it already acknowledges that some domains get removed after review. If you hard-reject on the deny list without ever consulting the allow list, you are ignoring the maintainers’ own correction mechanism.

FakeFilter publishes an automated list of temp-mail providers daily and offers an online API at fakefilter.net. That daily cadence is useful for catching new providers fast, and it is also a reason to be cautious: automated scraping produces churn, and domains get added before they are fully vetted. A match from a fast-moving automated feed deserves less weight than a match from a list with manual review. Your rejection logic should reflect that difference.

MailChecker supports over 55,000 throwable email domains, which is excellent coverage and also a reminder of scale. The larger the list, the more likely it contains domains that are only tangentially related to disposable email, such as catch-all domains used by small businesses or regional ISPs. Coverage and precision pull in opposite directions. Build your pipeline assuming that trade-off exists.

Log the matched source before you log the rejection

This is the single highest-leverage change you can make. When a check fires, record the domain, the list name, the list version or fetch timestamp, and the match type. Without that metadata you cannot tell whether your false-positive rate comes from one noisy feed or is spread evenly. You also cannot answer a support ticket that says “why was I blocked” without digging through logs that probably do not exist.

Source logging also lets you weight lists differently over time. If you notice that one feed accounts for 80 percent of verified-user complaints, you can demote it to a soft signal without rewriting your whole pipeline. The burner-email-providers repository is a plain list of temporary providers, useful as a signal but not as an authority. Knowing which list produced a hit is what turns that distinction into working code.

Build a tiered response instead of a binary gate

A tiered response means different match conditions produce different outcomes. A domain that appears on three independent lists with recent timestamps is a strong signal. A domain that appears on one automated feed and nowhere else is a weak signal. The weak signal should trigger a confirmation email, not a lockout. The user proves control of the address, and you get a verified signup either way.

This matters most at the margins. A legitimate user on a flagged domain can verify via a confirmation email and proceed. A burner user usually cannot or will not, because the whole point of the throwaway address is to skip verification. Verification does the filtering work that hard rejection was trying to do, without the collateral damage. The ivolo/disposable-email-domains repository is no longer maintained, which is a good example of why you should not build a permanent rejection around any single source. Lists go stale. Verification does not.

The amieiro/disposable-email-domains project regenerates aggregated deny and allow lists every 15 minutes. That refresh rate is great for coverage and terrible for stability if you cache aggressively or reject on first sight. A domain can move between lists within a single signup session, and allmail.one/blog/disposable-email-domain-lists shows how such shifts happen. Design your check so a transient match does not become a permanent account flag.

Route low-confidence matches to verification, not rejection

Set three bands: allow, verify, and block. Allow means the domain is on a trusted allow list or has no matches. Verify means at least one list matched but confidence is below your block threshold. Block means multiple independent sources agree and the timestamps are recent. Most real-world traffic should land in allow or verify. If your block band is catching more than a fraction of a percent of signups, your thresholds are too aggressive.

DeBounce provides a free disposable-check endpoint without an API key, which makes it easy to prototype the verify band quickly. Kickbox and similar email validation APIs offer disposable checks alongside syntax and MX validation, so you can combine signals in one call. The point is not to pick one vendor. The point is to have more than one input before you decide a user is not worth keeping.

Source Model Refresh cadence Best use in a tiered pipeline
disposable-email-domains Community blocklist plus allowlist, CC0 licensed Community-driven, irregular Primary deny signal, always cross-check the allowlist
FakeFilter Automated list plus API Daily Fast catch of new providers, treat as medium confidence
MailChecker Library with 55,000+ domains Per release Bulk matching in-app, good for verify band
amieiro/disposable-email-domains Aggregated deny and allow lists Every 15 minutes Fresh aggregation, avoid caching too long
DeBounce Hosted API, free endpoint Vendor-managed Secondary opinion before blocking

Keep the allow list in the loop at all times

An email domain blocklist without an override path is a liability. Domains change hands, providers clean up their act, and shared infrastructure gets reallocated. If your only response to a match is rejection, you have no mechanism to correct a wrong call. The allow list is that mechanism, and it should be consulted before any block decision, not after a user complains.

The disposable-email-domains project ships an allowlist alongside its blocklist for exactly this reason. Respect it. If a domain appears on both, the allow list should win or at minimum downgrade the match to the verify band. This is not a loophole. It is the maintainers telling you that a domain was reviewed and cleared.

AllMail publishes a guide on disposable email domain lists that covers how these sources differ and how to combine them. Reading up on the differences before you wire three feeds into a single boolean is worth an afternoon. Most teams that get burned by false positives did not lack sources. They lacked a policy for resolving conflicts between sources.

Monitor blocklist removal and false-positive rates as first-class metrics

Track how often a blocked domain later gets removed from the list that flagged it. That number is your false-positive proxy. If it is climbing, your block threshold is too tight or one of your feeds is noisy. Pair it with a simple metric: signups that failed verification on a flagged domain but succeeded when retried on a different address. That tells you how many real users you almost lost.

Set up blocklist removal monitoring so you know when a domain drops off a list you depend on. If you keep a local cache and never invalidate it, a domain can stay blocked in your system long after it was cleared upstream. The amieiro project’s 15-minute regeneration is a reminder that lists are living data. Your cache TTL should be shorter than you think.

Open-source email lists make this easy because you can diff versions. Pull the raw files from GitHub, store a hash, and alert on changes to domains you have blocked. That is a few hours of work and it eliminates a whole class of support tickets.

Start by adding source logging to every check today, even before you change any thresholds. Then split your decision into allow, verify, and block, and route every single-source match to verification. Keep the allow list in the decision path, watch removal rates, and shorten your cache TTL. The goal is not to catch every burner email. It is to catch the ones that matter without locking out the people who can prove they are real.