Bot protection and anti-enumeration

A bot testing addresses in series on a form leaves a trace: the very shape of what it submits. YesWeCheck analyses four signals on the flow of each organisation, recognises three attack families, and warns by email with one-click blocking. With no configuration.

What a bot is after

An address validation form answers a useful question, and that is what makes it worth hijacking. Submitted thousands of times, it becomes a discovery tool: the attacker learns which addresses exist on a domain, without ever having had to guess right the first time.

The pattern cannot be seen on an isolated request. Each submission is perfectly ordinary: an address, a domain, a response. What gives the attack away is the relationship between submissions, on the same domain, within the same interval.

The four signals

Each one measures a form of regularity that a machine produces and that a run of human sign-ups does not. None is enough on its own: it is their combination on the same domain, within the same interval, that tells an automated campaign apart from an ordinary day of sign-ups.

The 4 signals analysed on the flow of each organisation.
SignalWhat it measures
AAverage Levenshtein distance between the local parts submitted on the same domain.
BNumeric progression on a common root, such as a name followed by a counter that keeps increasing.
CConcentration of random local parts on the same domain, the signature of a list bought in bulk.
DCollapse in n-gram diversity, the signature of guessing by permuting first and last names.

The thresholds are not published, by choice: publishing the exact point at which an alert fires would amount to handing over the instructions for staying just below it.

The three attack families

The signals group into three ways of going about it, which match three different intentions. Recognising which one is under way changes what the organisation has to do: the first is blocked, the second is filtered upstream, the third targets one specific person.

The 3 attack families recognised.
FamilySignalsWhat happens
Sequential enumerationA, BA bot works through a form, testing a run of addresses close to one another.
Bulk list injectionCRandom addresses arrive in large numbers on the same domain, recent or disposable.
Permutation guessingDEvery combination of the same person's first and last name is tried, one after the other.

Detection is silent

Detection is silent: no block is returned to the caller, so the attacker never learns they have been spotted. The organisation receives an email alert, with concrete examples and one-click blocking. The block therefore falls when the organisation decides, not at a moment the attacker could notice.

It is a design choice, and it runs against intuition. Returning an error the moment the attack is spotted would be satisfying, and it would be information handed to the attacker: they would know straight away which pace and which shape of submission trigger detection, and therefore how to escape it.

The API therefore keeps answering normally. It is the organisation that is warned, outside the channel the attacker is watching, and it is the organisation that decides when the block falls.

Where the protection applies

The three entry surfaces are covered in the same way, with nothing to set up. An organisation that validates its addresses from its server, from a public form and from an imported file benefits from the same analysis: it is the flow that is observed, not the entry point.

See what protects a validation widget

Frequently asked questions

What is email address enumeration?
It is the act of submitting addresses in series to a form to find out which ones are accepted. The attacker is not trying to sign up: they use the form response as a revealer, to rebuild a list of real addresses on a domain.
Why does detection stay silent?
Detection is silent: no block is returned to the caller, so the attacker never learns they have been spotted. The organisation receives an email alert, with concrete examples and one-click blocking. The block therefore falls when the organisation decides, not at a moment the attacker could notice.
Is there anything to configure?
No. Protection is active on the API, on the widget and on batch processing, with no setting. There is no threshold to choose and no rule to write: the analysis runs on the real flow of each organisation, and the alert goes out when a pattern takes shape.
Does this protection add to a captcha?
Yes, the two complement each other. A captcha filters before submission; this detection reads what has already been submitted, without asking the visitor anything. The YesWeCheck widget also offers its own computing challenge, run in the browser, with no image to decipher.

Go further

Enumeration and the injection of purchased lists overlap with two other checks: the randomly drawn local part and the disposable domain. A random address on a temporary domain carries both signals at once, and the response says so check by check.

Content updated on 7 September 2026.