---
title: "Refuse password guessing from known addresses | FoxAuth"
description: "Use FoxTrust’s signed list of addresses seen guessing passwords: count sign-ins from them first, then refuse their passwords project by project."
canonical: "https://foxauth.dev/docs/get-started/refuse-password-guessing-addresses/"
lastmod: "2026-11-10T00:23:46+03:00"
---
Refuse password guessing from known addresses

Someone who holds a leaked password list tries it against your accounts from one address, many accounts at a time. The
sign-in throttle counts wrong passwords per account, so an address that tries each account only a few times is not
stopped by it. This page shows how a project can refuse password sign-ins from addresses that are known to do this.

What the list holdsSection titled “What the list holds”

FoxTrust is the reputation service that receives FoxAuth’s attack reports. It publishes a list of the addresses it saw
guessing passwords in the last 7 days. It rebuilds the list every hour and signs it.

Today the list holds about 3,000 addresses. Most of them come from third-party sources.

FoxTrust never lists an address because of one server’s reports alone.

An IPv4 entry is one address. An IPv6 entry is a /64 network, so it covers every address in that network.

A shared list holds shared addresses: a mobile operator’s, an office’s, a public Wi-Fi’s. That is why each project
counts first and refuses only when its own numbers say so.

What FoxTrust learnsSection titled “What FoxTrust learns”

Nothing about who signs in. Your server downloads the list and checks every sign-in against it on its own. The
download sends FoxTrust the name “FoxAuth” and, while usage reporting is on, the FoxAuth version. It sends no name of
your server, no address of a person and no account.

The address of a refused person is not written anywhere: not in a log line, an event, an error record, the audit trail
or a counter. It appears only on the page shown to that person.

Turn the list onSection titled “Turn the list on”

A super administrator opens Settings → Integrations and, under Attack reporting, turns on Use FoxTrust’s list
of known password-guessing addresses. It is off by default.

The switch needs Report attacking addresses to FoxTrust on, because the list is built from what servers report.
Turning attack reporting off later asks first and says how many projects count or refuse with the list. Once confirmed, those projects
refuse nobody.

From then on, each server process:

downloads the list every hour, shortly after FoxTrust rebuilds it, and within a minute of the switch going on;

checks the signature against a key built into FoxAuth, and refuses a list that is older than the one it holds, or
that has expired;

uses a list for at most 3 hours after FoxTrust built it.

The card under the switch shows when the list was built, how many addresses it holds, until when it is valid, and the
last download that was rejected, with the reason.

Decide with Count onlySection titled “Decide with Count only”

Each project chooses on its Sign-in tab, in the card Password guessing from known addresses:

Mode
What happens

Nothing
Sign-ins from listed addresses are neither counted nor refused.

Count only
The default once the list is on. Nothing changes for anyone. The project counts.

Refuse password sign-in
A listed address that keeps failing passwords is refused, as described below.

The card shows three numbers for the last 7 days:

Password sign-ins from listed addresses;

Of those, with the right password;

Refused because of the list.

Leave the project on Count only for a while. If many sign-ins from listed addresses had the right password, your own
people share addresses with attackers, for example behind an office’s or a mobile operator’s address. Add those
networks to the exemptions below before you refuse. If almost none had the right password, refusing is safe.

What Refuse password sign-in doesSection titled “What Refuse password sign-in does”

Choosing Refuse password sign-in shows the project’s numbers again and asks you to confirm. Then:

An address on the list, seen guessing in the last 48 hours, is refused once it has failed 3 passwords on this server
within an hour.

An address that FoxTrust confirms from independent sources is refused at once, without those 3 failures.

A refused address stays refused while it keeps trying. After an hour without attempts it gets 3 attempts again.

A person on a listed address who types the right password first signs in as before.

A refused attempt is not checked at all: the password is not compared and the account is not looked up. It does not
count against the account’s own sign-in throttle, so an attacker cannot lock the owner out this way.

The 3 failures are counted per server, not per project: an address that guesses passwords in one project can be
refused in another project that chose Refuse. On a server with several processes, each process counts its own.

Only password sign-in is affected. Sign-in through an identity provider, password reset, sign-up and the token
endpoints are not. Refuse password sign-in is not offered for the administrators’ project, because it could lock
every administrator out of the console.

What a refused person seesSection titled “What a refused person sees”

The password form is replaced by a short explanation, in the person’s language (English, Russian or Chinese):

Password sign-in is not available from this network.

What happened: this network’s address was recently used to guess passwords.

What to do now: try again from another network, for example mobile data instead of Wi-Fi, or use one of the other
ways to sign in on the page. The project’s identity provider buttons stay.

A line to send to the people who run the service, with their own address and the time in UTC:

listed-address · 203.0.113.7 · 2026-11-09 19:58 UTC

Last, a link: Ask FoxTrust to review this address.

The page is the same whatever email was typed, and whether or not the account exists.

Help a refused personSection titled “Help a refused person”

When someone sends you that line, add the address from it under Never check these networks on the project’s
Sign-in tab. The person can sign in again at once. Since the address is never logged, their line is the only place
it comes from.

Never check these networksSection titled “Never check these networks”

A project can list addresses or ranges that are never checked against the list, such as an office or a VPN’s exit
address. Each one can end on a date, after which the list applies to it again. Nothing is counted for an exempt
address.

Write a single address, such as 203.0.113.7, or a range, such as 198.51.100.0/24. IPv6 works the same way.

Your address now fills in the address you are using, so you can exempt your own network in one step.

A project holds at most 100 exemptions.

A wrong listingSection titled “A wrong listing”

If an address is on the list by mistake, ask FoxTrust to review it on its dispute page.
The link on the refusal page leads there too. Until the listing is removed, exempt the network as above.

When the list is out of dateSection titled “When the list is out of date”

A list that is missing, more than 3 hours old or in a format this FoxAuth cannot read refuses nobody. Every project that
uses the list says so on its Sign-in tab, for example: “The list is out of date, so nobody is refused because of it.” This happens
when FoxTrust cannot be reached for a few hours. Protection resumes with the next list that verifies.

Note

A server whose caller address is set up wrong sees every person as its own
proxy’s address. The list is never checked for a trusted proxy’s address, so
such a server refuses nobody rather than everybody. The attack reporting card
warns about it.

For an AI agentSection titled “For an AI agent”

An agent connected to the administrative MCP endpoint does the same, through the console’s own routes and audit trail:
attacker_list_status_get reads the state of the list, project_password_guessing_get reads a project’s mode,
exemptions and numbers, and project_update with passwordGuessing changes them. Turning the list on is the instance
setting attackerList.enabled.

PreviousAdd conditions at your proxyNextLet people turn on two-step verification
