13 Commits

Author SHA1 Message Date
Dan Milne
209c5496d8 Fix asset precompile boot and bump version to 0.16.0
Some checks failed
CI / scan_ruby (push) Has been cancelled
CI / scan_js (push) Has been cancelled
CI / scan_container (push) Has been cancelled
CI / lint (push) Has been cancelled
CI / test (push) Has been cancelled
CI / system-test (push) Has been cancelled
The CLINCH_HOST initializer raised during `assets:precompile` in the
Docker build, where no real host is set. Skip the check when
SECRET_KEY_BASE_DUMMY is present (the build-time precompile step);
deployed boots still require CLINCH_HOST.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-11 23:53:09 +10:00
Dan Milne
d49e7ce4f5 Move CSP to nonces; remove unsafe-inline from script-src and style-src
unsafe-inline on script-src neutered CSP as an XSS defense on the login and OAuth
consent pages (the highest-value targets in an IdP). Switch to a per-response
nonce for both script-src and style-src and drop unsafe-inline entirely.

- Add a random per-response nonce generator and apply it to script-src/style-src.
- Remove :unsafe_inline from both directives.
- Nonce the one hand-written inline script (dark-mode detection in the layout).
- Convert the 2 static style="display:none" attributes to class="hidden" (their
  runtime toggle is done via element.style in JS, which CSP does not govern).

importmap-rails (2.2.3) already stamps the nonce onto its generated inline
importmap/module/preload tags, and Turbo (2.0.23) reads csp_meta_tag for its
injected <style>, so no other view changes were needed. Adds an integration test
asserting the enforcing header carries nonces, omits unsafe-inline, and that the
inline script's nonce matches the header.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 20:42:28 +10:00
Dan Milne
44892e3301 Make WebAuthn clone detection actually block, and fix false positives
Two problems with sign-count clone detection:

- suspicious_sign_count? flagged the case where both the stored and presented
  counts are 0. Most synced passkeys (Apple/Google) report 0 every time, so every
  legitimate sign-in was flagged — drowning real signals in noise. Per WebAuthn
  §6.1.1 a 0 counter means "no counter"; only flag when BOTH counts are non-zero
  and the new one does not advance.

- On a suspicious count the controller only logged a warning and then continued
  to authenticate and overwrite the stored counter. A cloned credential therefore
  worked indefinitely. webauthn_verify now rejects the sign-in (no session, no
  counter update) and emails the user via a new SecurityMailer#suspicious_passkey_used.

Tests cover the corrected classification (synced/first-use/normal vs equal/
decreasing) and the new alert email.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 20:28:38 +10:00
Dan Milne
24266872f9 Revoke access tokens too on refresh-token reuse detection
revoke_family! revoked only the refresh tokens in a rotation family. When reuse
of a revoked refresh token was detected (a token-theft signal), the access
tokens issued across that chain stayed valid at /userinfo until expiry — up to
the access-token TTL — so an attacker holding a stolen access token kept access.

revoke_family! now also revokes every access token referenced by the family's
refresh tokens. Adds a regression test: rotate once, reuse the revoked token,
and assert both the original and rotated-in access tokens are revoked.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 20:23:17 +10:00
Dan Milne
cd862c7cd7 Filter code params from logs (TOTP, backup, OAuth code, PKCE)
The TOTP and backup-code form field is named `code`, which was not covered by
the filter list, so live one-time codes landed in production logs. Adding :code
(partial match) also redacts the OAuth authorization `code` and PKCE
`code_verifier`/`code_challenge`.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 20:21:41 +10:00
Dan Milne
89bd5f1432 Enforce account-active status across the auth lifecycle
active? was only checked at the password step of sign-in. A user disabled
afterwards could (a) still complete the 2FA step and mint a valid session, and
(b) keep using any existing session until natural expiry, because per-request
auth only checked session expiry, not user status.

Three enforcement points:
- Mid-flow guard: verify_totp and webauthn_verify re-check active? before
  start_new_session_for, clearing the pending session and rejecting if disabled.
- Request-time guard: find_session_by_cookie now uses Session.for_active_user,
  so a session whose user is disabled no longer authenticates (authoritative,
  catches any disable path including direct DB changes).
- Immediate cleanup: User#revoke_sessions_when_deactivated destroys a user's
  sessions when status changes away from active, so access is revoked everywhere
  at once rather than on the next request.

Tests cover the mid-flow TOTP rejection, request-time rejection of an existing
session after disable, session destruction on disable, and that unrelated
updates leave sessions intact.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 19:53:50 +10:00
Dan Milne
57d7d1f691 Anchor host-authorization regex to prevent look-alike domain bypass
The DNS-rebinding allowlist used /.*#{registrable_domain}/, which is unanchored:
for example.com it also matched evil-example.com, notexample.com,
example.computer, and example.com.attacker.com. Any of those hosts would pass
Rails' HostAuthorization middleware.

Anchor the pattern as /\A(.+\.)?DOMAIN\z/i so it matches only the registrable
domain and its subdomains (now also case-insensitively). Verified against a
real production boot.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 19:47:35 +10:00
Dan Milne
406a79d9eb Block SSRF via backchannel_logout_uri
backchannel_logout_uri was validated only for scheme/HTTPS, so an admin (or a
compromised admin account) could point it at internal infrastructure — cloud
metadata (169.254.169.254), loopback, or RFC1918 hosts — and every user logout
would fire a server-side POST there.

Add PrivateAddressCheck (app/lib) and apply it as defense-in-depth:
- Application validation rejects URIs whose host is, or is a literal, internal
  address (loopback / private / link-local / 0.0.0.0 / localhost / metadata
  hostnames). Fast, DNS-free, immediate admin feedback.
- BackchannelLogoutJob re-checks at request time WITH DNS resolution and aborts
  (no retry) if the host resolves to a non-public address — covering URIs that
  predate the validation and public hostnames pointed at internal IPs.

Tests cover the address classification, the model validation, and updates an
existing test that used a localhost logout URI.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 08:14:45 +10:00
Dan Milne
f38ac2ecc8 Prevent TOTP code replay within the drift window
verify_totp called ROTP without `after:`, so a captured 6-digit code stayed
valid for the full ~90s drift window and could be replayed in a separate
sign-in. Add a last_otp_at column, pass it as ROTP's `after:`, and persist the
matched timestep on success so a code (or any earlier one) cannot be reused.

Also fixes a latent bug surfaced by the new replay path: enable_totp! did
`self.backup_codes = generate_backup_codes`, reassigning backup_codes to the
plaintext return value (generate_backup_codes already stores the BCrypt hashes
internally). That stored backup codes in plaintext and broke verification.
enable_totp! is test-only today, but it is public and backup_codes is not
encrypted, so this is a real footgun. Now it just calls generate_backup_codes.

Rewrites the mislabeled "TOTP code cannot be reused" test to actually assert
that replaying an accepted code is rejected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 08:10:34 +10:00
Dan Milne
84ed462f40 Require CLINCH_HOST in deployed environments; drop request-host fallback
determine_base_url fell back to request.host when CLINCH_HOST was unset. Rails
resolves request.host from X-Forwarded-Host behind a trusted proxy, so a spoofed
header could make the forward-auth login redirect point at an attacker origin
(host-header phishing).

- Add config/initializers/clinch_host.rb: fail fast at boot in any non-local
  environment when CLINCH_HOST is blank. It anchors the OIDC issuer, WebAuthn
  RP ID, and login redirect, so it must be explicit, never inferred.
- determine_base_url now uses CLINCH_HOST (guaranteed in production) with a safe
  localhost default for dev/test, and never reads the request host.
- Simplify the spoofed-host regression test now that the fallback is safe.

Verified: production boot aborts with a clear message when CLINCH_HOST is blank,
and boots normally when set.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 08:04:42 +10:00
Dan Milne
96a657e349 Validate X-Forwarded-Host before using it as a post-login redirect target
render_unauthorized built the post-login return URL directly from the
attacker-influenceable X-Forwarded-Host / X-Forwarded-Uri headers, stored it
in the session, and reflected it into the signin `rd`. After authentication
that URL is followed with allow_other_host, so a spoofed host was an open
redirect.

Now the forwarded URL is only honoured if it resolves to a known, active
forward-auth application (via validate_redirect_url); otherwise it falls back
to a validated `rd` or the IdP's base URL. Once render_unauthorized only ever
stores a validated value, the sessions_controller "supplement, don't replace"
behaviour is safe, so no change is needed there.

Two integration tests were asserting the old behaviour by reflecting
unregistered hosts (grafana.example.com, app.example.com); they now register
those domains as forward-auth apps so they exercise the real feature. Adds a
regression test that a spoofed X-Forwarded-Host is neither stored nor
reflected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 08:00:12 +10:00
Dan Milne
8a095e4939 Enforce group access on Bearer API key forward-auth at use-time
The ApiKey model only validates group access on creation (user_must_have_access
runs on create). The bearer path in /api/verify never re-checked, so a user
removed from an application's allowed groups kept access via an existing key
until it was manually revoked.

Add an app.user_allowed?(user) check to authenticate_bearer_token, matching the
session path, returning 401 when the user no longer has group access. Adds a
regression test that revokes membership after key creation.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 07:54:48 +10:00
Dan Milne
703d24e4e4 Fix ForwardAuth fail-open and consent CSRF bypass
Two HIGH-severity findings from the security review:

- ForwardAuth: when no host header was present, /api/verify skipped the
  application lookup and group check entirely, returning 200 with identity
  headers (including all of the user's groups). This bypassed per-domain
  access control. Now fails closed with 403, and the unreachable
  DEFAULT_HEADERS fallback (the bypass path) is removed so headers are
  always scoped to a resolved, active application.

- OIDC: the consent endpoint was in the verify_authenticity_token skip
  list, so a forged cross-site POST could silently grant OAuth scopes.
  Removed :consent from the skip list (the form already embeds the token).

Adds regression tests for both: fail-closed with no identity headers when
host is absent, and 422 on a tokenless consent POST.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-11 07:52:56 +10:00
37 changed files with 627 additions and 69 deletions

View File

@@ -64,26 +64,16 @@ module Api
return render_forbidden("No authentication rule configured for this domain")
end
else
Rails.logger.info "ForwardAuth: User #{user.email_address} authenticated (no domain specified)"
end
headers = if app
app.headers_for_user(user)
else
Application::DEFAULT_HEADERS.map { |key, header_name|
case key
when :user, :email, :name
[header_name, user.email_address]
when :username
[header_name, user.username] if user.username.present?
when :groups
user.groups.any? ? [header_name, user.groups.map(&:name).join(",")] : nil
when :admin
[header_name, user.admin? ? "true" : "false"]
end
}.compact.to_h
# Fail closed: with no host we cannot resolve an application or evaluate its
# group policy. Emitting identity headers here would bypass all per-domain
# access control, so reject instead.
Rails.logger.info "ForwardAuth: Access denied - no host header present"
return render_forbidden("No host header present")
end
# Reaching here implies a matching, active application was resolved above
# (every other path returns forbidden), so headers are always scoped to it.
headers = app.headers_for_user(user)
headers.each { |key, value| response.headers[key] = value }
Rails.logger.debug "ForwardAuth: Headers sent: #{headers.keys.join(", ")}" if headers.any?
@@ -148,6 +138,14 @@ module Api
return render_bearer_error("Application is inactive")
end
# Re-check group membership at use-time. The ApiKey model only validates
# access on creation, so a user removed from the app's allowed groups
# afterwards must not keep access via an existing key.
unless app.user_allowed?(user)
Rails.logger.info "ForwardAuth: API key '#{api_key.name}' denied - user #{user.email_address} lacks group access to #{app.domain_pattern}"
return render_bearer_error("Access denied: insufficient group membership")
end
api_key.touch_last_used!
headers = app.headers_for_user(user)
@@ -198,11 +196,14 @@ module Api
original_host = request.headers["X-Forwarded-Host"]
original_uri = request.headers["X-Forwarded-Uri"] || request.headers["X-Forwarded-Path"] || "/"
original_url = if original_host
"https://#{original_host}#{original_uri}"
else
redirect_url || base_url
end
# X-Forwarded-Host is attacker-influenceable, so only honour the forwarded
# URL as a post-login redirect target if it resolves to a known, active
# forward-auth application. Otherwise this is an open redirect: a spoofed
# host would be stored and reflected into the signin `rd`, then followed
# (with allow_other_host) after the user authenticates. Fall back to a
# validated `rd` or, failing that, the IdP's own base URL.
forwarded_url = "https://#{original_host}#{original_uri}" if original_host.present?
original_url = validate_redirect_url(forwarded_url) || redirect_url || base_url
session[:return_to_after_authenticating] = original_url
@@ -242,18 +243,13 @@ module Api
def determine_base_url(redirect_url)
return redirect_url if redirect_url.present?
if ENV["CLINCH_HOST"].present?
host = ENV["CLINCH_HOST"]
host.match?(/^https?:\/\//) ? host : "https://#{host}"
else
request_host = request.host || request.headers["X-Forwarded-Host"]
if request_host.present?
Rails.logger.warn "ForwardAuth: CLINCH_HOST not set, using request host: #{request_host}"
"https://#{request_host}"
else
raise StandardError, "ForwardAuth: CLINCH_HOST environment variable not set and no request host available."
end
end
# CLINCH_HOST is the IdP's canonical origin and is mandatory in deployed
# environments (enforced at boot in config/initializers/clinch_host.rb).
# We never fall back to the request host: a spoofed X-Forwarded-Host would
# otherwise redirect the login flow to an attacker-controlled origin. The
# localhost default only applies to local dev/test.
host = ENV["CLINCH_HOST"].presence || "http://localhost:3000"
host.match?(%r{\Ahttps?://}) ? host : "https://#{host}"
end
end
end

View File

@@ -31,7 +31,7 @@ module Authentication
end
def find_session_by_cookie
Session.active.find_by(id: cookies.signed[:session_id]) if cookies.signed[:session_id]
Session.active.for_active_user.find_by(id: cookies.signed[:session_id]) if cookies.signed[:session_id]
end
def request_authentication

View File

@@ -4,7 +4,11 @@ class OidcController < ApplicationController
# Discovery and JWKS endpoints are public
# authorize is also unauthenticated to handle prompt=none and prompt=login specially
allow_unauthenticated_access only: [:discovery, :jwks, :token, :revoke, :userinfo, :logout, :authorize]
skip_before_action :verify_authenticity_token, only: [:token, :revoke, :userinfo, :logout, :authorize, :consent]
# Machine-to-machine endpoints (token/revoke/userinfo) and pure redirect handlers
# (logout/authorize) legitimately skip CSRF. The consent endpoint is browser-facing
# and state-changing (it grants OAuth scopes), so it MUST keep CSRF protection — the
# consent form already embeds the token via form_with.
skip_before_action :verify_authenticity_token, only: [:token, :revoke, :userinfo, :logout, :authorize]
# RFC 6749 §4.1.2.1: client_id and redirect_uri must be validated *before* any
# other error can be reported via redirect. Failures here render a plain page.

View File

@@ -121,6 +121,16 @@ class SessionsController < ApplicationController
return
end
# Re-check account status: active? was verified at the password step, but an
# admin may have disabled the account while the user sat on this 2FA screen.
# Without this, a disabled account could still mint a valid session here.
unless user.active?
session.delete(:pending_totp_user_id)
session.delete(:pending_remember_me)
redirect_to signin_path, alert: "Your account is not active. Please contact an administrator."
return
end
remember_me = session.delete(:pending_remember_me) || false
# Try TOTP verification first (password + TOTP = 2FA)
@@ -241,6 +251,14 @@ class SessionsController < ApplicationController
return
end
# Re-check account status: an admin may have disabled the account between the
# password step and this passkey verification. Reject before creating a session.
unless user.active?
session.delete(:pending_webauthn_user_id)
render json: {error: "Your account is not active."}, status: :unauthorized
return
end
# Get the credential and assertion from params
credential_data = params[:credential]
if credential_data.blank?
@@ -277,10 +295,14 @@ class SessionsController < ApplicationController
sign_count: stored_credential.sign_count
)
# Check for suspicious sign count (possible clone)
# Clone detection: a non-advancing signature counter signals the credential
# may have been copied. Reject the sign-in (do NOT create a session or update
# the stored counter) and alert the user, per WebAuthn §6.1.1.
if stored_credential.suspicious_sign_count?(webauthn_credential.sign_count)
Rails.logger.warn "Suspicious WebAuthn sign count for user #{user.id}, credential #{stored_credential.id}"
# You might want to notify admins or temporarily disable the credential
Rails.logger.warn "Suspicious WebAuthn sign count for user #{user.id}, credential #{stored_credential.id} (stored=#{stored_credential.sign_count}, presented=#{webauthn_credential.sign_count})"
SecurityMailer.suspicious_passkey_used(user, nickname: stored_credential.display_name, **security_event_context).deliver_later
render json: {error: "Passkey authentication could not be completed. Please contact support."}, status: :unprocessable_entity
return
end
# Update credential usage

View File

@@ -28,6 +28,14 @@ class BackchannelLogoutJob < ApplicationJob
# Send HTTP POST to the application's backchannel logout URI
uri = URI.parse(application.backchannel_logout_uri)
# SSRF guard: re-check at request time (with DNS resolution) in case the URI
# predates the validation, or a public hostname now resolves to an internal
# address. Abort without retrying — retries would not change the outcome.
if PrivateAddressCheck.internal_host?(uri.host) || PrivateAddressCheck.resolves_to_internal?(uri.host)
Rails.logger.error "BackchannelLogout: Refusing to send logout to #{application.name} - #{uri.host} is or resolves to a non-public address (SSRF guard)"
return
end
begin
response = Net::HTTP.start(uri.host, uri.port, use_ssl: uri.scheme == "https", open_timeout: 5, read_timeout: 5) do |http|
request = Net::HTTP::Post.new(uri.path.presence || "/")

View File

@@ -0,0 +1,57 @@
require "ipaddr"
require "resolv"
# SSRF guard for outbound requests to admin-configured URLs (currently the OIDC
# backchannel logout endpoint). Blocks hosts that are, or resolve to, private,
# loopback, link-local (incl. the cloud metadata address 169.254.169.254) or
# otherwise non-public address space.
module PrivateAddressCheck
module_function
# Hostnames that are internal by definition and must never be dialled.
BLOCKED_HOSTNAMES = %w[localhost metadata.google.internal].freeze
# Fast, DNS-free check: catches IP literals and well-known internal hostnames.
# Suitable for model validation (deterministic, immediate admin feedback).
def internal_host?(host)
host = host.to_s.downcase
return true if host.blank?
return true if BLOCKED_HOSTNAMES.include?(host)
return true if host.end_with?(".localhost")
ip = parse_ip(host)
ip ? internal_ip?(ip) : false
end
# Authoritative check: resolves the hostname and blocks if ANY address is
# internal. Suitable for request time — also defeats a public hostname that
# has been pointed at an internal IP (DNS rebinding to internal space).
def resolves_to_internal?(host)
addresses(host).any? { |ip| internal_ip?(ip) }
end
def addresses(host)
ip = parse_ip(host)
return [ip] if ip
Resolv.getaddresses(host.to_s).filter_map { |a| parse_ip(a) }
rescue StandardError
# Resolution failure: surface no addresses. Callers treat "can't resolve" as
# not-provably-internal; the dial itself will then fail safely.
[]
end
def internal_ip?(ip)
ip.loopback? || ip.private? || ip.link_local? || unspecified?(ip)
end
def parse_ip(str)
IPAddr.new(str.to_s)
rescue IPAddr::Error
nil
end
def unspecified?(ip)
ip == IPAddr.new("0.0.0.0") || ip == IPAddr.new("::")
end
end

View File

@@ -40,6 +40,12 @@ class SecurityMailer < ApplicationMailer
mail subject: "#{SUBJECT_PREFIX}An API key was revoked on your account", to: user.email_address
end
def suspicious_passkey_used(user, nickname:, ip:, user_agent:, occurred_at:)
assign_context(user, ip, user_agent, occurred_at)
@nickname = nickname
mail subject: "#{SUBJECT_PREFIX}A passkey sign-in was blocked", to: user.email_address
end
def email_address_changed(user, recipient:, old_email:, new_email:, ip:, user_agent:, occurred_at:)
assign_context(user, ip, user_agent, occurred_at)
@recipient = recipient

View File

@@ -56,6 +56,7 @@ class Application < ApplicationRecord
message: "must be a valid HTTP or HTTPS URL"
}
validate :backchannel_logout_uri_must_be_https_in_production, if: -> { backchannel_logout_uri.present? }
validate :backchannel_logout_uri_not_internal, if: -> { backchannel_logout_uri.present? }
# Icon validation using ActiveStorage validators
validate :icon_validation
@@ -390,4 +391,17 @@ class Application < ApplicationRecord
# Let the format validator handle invalid URIs
end
end
# SSRF guard: the backchannel logout URI is dialled server-side on every user
# logout, so it must not target internal infrastructure (loopback, private
# ranges, or the link-local cloud metadata endpoint). This is the fast,
# config-time check; BackchannelLogoutJob re-checks with DNS resolution.
def backchannel_logout_uri_not_internal
uri = URI.parse(backchannel_logout_uri)
if uri.host.present? && PrivateAddressCheck.internal_host?(uri.host)
errors.add(:backchannel_logout_uri, "must not point to a private, loopback, or link-local address")
end
rescue URI::InvalidURIError
# Let the format validator handle invalid URIs
end
end

View File

@@ -49,11 +49,21 @@ class OidcRefreshToken < ApplicationRecord
update!(revoked_at: Time.current)
end
# Revoke all refresh tokens in the same family (token rotation security)
# Revoke all refresh tokens in the same family (token rotation security).
# Also revoke every access token issued within the family: on a detected reuse
# attack the stolen chain's access tokens must not remain usable at /userinfo
# until they expire.
def revoke_family!
return unless token_family_id.present?
OidcRefreshToken.in_family(token_family_id).update_all(revoked_at: Time.current)
now = Time.current
family = OidcRefreshToken.in_family(token_family_id)
access_token_ids = family.pluck(:oidc_access_token_id).compact.uniq
family.update_all(revoked_at: now)
if access_token_ids.any?
OidcAccessToken.where(id: access_token_ids, revoked_at: nil).update_all(revoked_at: now)
end
end
private

View File

@@ -7,6 +7,9 @@ class Session < ApplicationRecord
# Scopes
scope :active, -> { where("expires_at > ?", Time.current) }
scope :expired, -> { where("expires_at <= ?", Time.current) }
# Sessions whose owning user is currently active. Used at request time so a
# disabled account cannot continue to authenticate with an existing session.
scope :for_active_user, -> { joins(:user).where(users: {status: User.statuses[:active]}) }
def expired?
expires_at.present? && expires_at <= Time.current

View File

@@ -41,6 +41,11 @@ class User < ApplicationRecord
# Enum - automatically creates scopes (User.active, User.disabled, etc.)
enum :status, {active: 0, disabled: 1, pending_invitation: 2}
# When an account stops being active (e.g. an admin disables it), immediately
# terminate its sessions so access is revoked everywhere, not just on expiry.
# Defence-in-depth: session lookup also filters by active status at request time.
after_update_commit :revoke_sessions_when_deactivated
# Scopes
scope :admins, -> { joins(:groups).where(groups: {admin: true}).distinct }
@@ -63,7 +68,10 @@ class User < ApplicationRecord
def enable_totp!
require "rotp"
self.totp_secret = ROTP::Base32.random
self.backup_codes = generate_backup_codes
# generate_backup_codes assigns the BCrypt hashes to self.backup_codes and
# returns the plaintext codes for display. Do NOT reassign backup_codes to the
# return value — that would store the plaintext codes and break verification.
generate_backup_codes
save!
end
@@ -86,7 +94,13 @@ class User < ApplicationRecord
require "rotp"
totp = ROTP::TOTP.new(totp_secret)
totp.verify(code, drift_behind: 30, drift_ahead: 30)
# Pass `after:` so a code can only be accepted once: ROTP rejects any timestep
# at or before the last accepted one, closing the ~90s drift-window replay.
verified_at = totp.verify(code, drift_behind: 30, drift_ahead: 30, after: last_otp_at)
return false unless verified_at
update_column(:last_otp_at, verified_at)
true
end
# Console/debug helper: get current TOTP code
@@ -237,6 +251,13 @@ class User < ApplicationRecord
Group.auto_assign.each { |g| groups << g }
end
def revoke_sessions_when_deactivated
return unless saved_change_to_status?
return if active?
sessions.destroy_all
end
def no_reserved_claim_names
return if custom_claims.blank?

View File

@@ -52,13 +52,17 @@ class WebauthnCredential < ApplicationRecord
end
end
# Check if sign count is suspicious (clone detection)
# Check if sign count is suspicious (clone detection).
#
# Per WebAuthn §6.1.1, a signature counter of 0 means the authenticator does
# not implement a counter (true of most synced passkeys — Apple/Google report
# 0 every time), so it cannot be used for clone detection. Only when BOTH the
# stored and presented counts are non-zero does a non-increasing value signal
# a possible clone.
def suspicious_sign_count?(new_sign_count)
return false if sign_count.zero? && new_sign_count > 0 # First use
return false if new_sign_count > sign_count # Normal increment
return false if sign_count.zero? || new_sign_count.zero?
# Sign count didn't increase - possible clone
true
new_sign_count <= sign_count
end
# Format for display in UI

View File

@@ -9,7 +9,7 @@
<%= csrf_meta_tags %>
<%= csp_meta_tag %>
<script>
<script nonce="<%= content_security_policy_nonce %>">
(function() {
var theme = localStorage.getItem('theme');
if (theme === 'dark' || (!theme && window.matchMedia('(prefers-color-scheme: dark)').matches)) {

View File

@@ -0,0 +1,16 @@
<p>Hello,</p>
<p>
A sign-in to your Clinch account (<strong><%= @user.email_address %></strong>)
using your passkey (<strong><%= @nickname %></strong>) was <strong>blocked</strong>
because its security counter did not advance as expected. This can indicate the
passkey has been copied (cloned).
</p>
<p>
If this was you and you are unable to sign in, remove this passkey and register
a new one. If you do not recognise this activity, treat it as a compromise:
remove the passkey and review your account security.
</p>
<%= render "event_metadata" %>

View File

@@ -0,0 +1,11 @@
Hello,
A sign-in to your Clinch account (<%= @user.email_address %>) using your passkey
("<%= @nickname %>") was BLOCKED because its security counter did not advance as
expected. This can indicate the passkey has been copied (cloned).
If this was you and you are unable to sign in, remove this passkey and register a
new one. If you do not recognise this activity, treat it as a compromise: remove
the passkey and review your account security.
<%= render "event_metadata" %>

View File

@@ -38,7 +38,7 @@
</svg>
Continue with Passkey
</button>
<div data-webauthn-target="error" class="mt-2 text-sm text-red-600" style="display: none;"></div>
<div data-webauthn-target="error" class="mt-2 text-sm text-red-600 hidden"></div>
</div>
<!-- Password section - shown by default, hidden if WebAuthn is required -->

View File

@@ -54,7 +54,7 @@
</svg>
Use Passkey Instead
</button>
<div data-webauthn-target="error" class="mt-2 text-sm text-red-600" style="display: none;"></div>
<div data-webauthn-target="error" class="mt-2 text-sm text-red-600 hidden"></div>
</div>
</div>
<% end %>

View File

@@ -118,14 +118,17 @@ Rails.application.configure do
registrable_domain = domain.domain # Gets "example.com" from "auth.example.com"
if registrable_domain.present?
# Create regex to allow any subdomain of the registrable domain
allowed_hosts << /.*#{Regexp.escape(registrable_domain)}/
# Allow the registrable domain and any subdomain of it. The pattern is
# anchored (\A...\z) with a mandatory dot before the domain so that
# look-alikes such as "evil-example.com" or "example.com.attacker.com"
# do NOT match — an unanchored /.*example\.com/ would allow both.
allowed_hosts << /\A(.+\.)?#{Regexp.escape(registrable_domain)}\z/i
end
rescue PublicSuffix::DomainInvalid
# Fallback to simple domain extraction if PublicSuffix fails
Rails.logger.warn "Could not parse domain '#{host_domain}' with PublicSuffix, using fallback"
base_domain = host_domain.split(".").last(2).join(".")
allowed_hosts << /.*#{Regexp.escape(base_domain)}/
allowed_hosts << /\A(.+\.)?#{Regexp.escape(base_domain)}\z/i
end
end

View File

@@ -0,0 +1,16 @@
# CLINCH_HOST is this IdP's canonical external origin, e.g. https://auth.example.com.
# It anchors the OIDC issuer, the WebAuthn RP ID, and the forward-auth login
# redirect. In deployed (non-local) environments it MUST be set explicitly and
# never inferred from request headers — X-Forwarded-Host is attacker-influenceable,
# so inferring the origin from it would allow host-header phishing and open
# redirects. Fail fast at boot rather than start in an unsafe configuration.
#
# Skipped during asset precompilation (e.g. the Docker build step, which sets
# SECRET_KEY_BASE_DUMMY): no real CLINCH_HOST exists yet and assets don't need it.
unless Rails.env.local? || ENV["SECRET_KEY_BASE_DUMMY"].present?
if ENV["CLINCH_HOST"].blank?
raise "CLINCH_HOST must be set (e.g. https://auth.example.com). It is the " \
"canonical origin of this Clinch instance and must not be inferred " \
"from request headers."
end
end

View File

@@ -9,13 +9,15 @@ Rails.application.configure do
# Default to self for everything, plus blob: for file downloads
policy.default_src :self, "blob:"
# Scripts: Allow self, importmaps, unsafe-inline for Turbo/StimulusJS, and blob: for downloads
# Note: unsafe_inline is needed for Stimulus controllers and Turbo navigation
policy.script_src :self, :unsafe_inline, "blob:"
# Scripts: self + per-response nonce (see nonce config below) + blob: for
# downloads. No unsafe-inline — importmap/Turbo/Stimulus inline tags carry the
# nonce automatically, and the one hand-written inline script is nonced.
policy.script_src :self, "blob:"
# Styles: Allow self and unsafe_inline for TailwindCSS dynamic classes
# and Stimulus controller style manipulations
policy.style_src :self, :unsafe_inline
# Styles: self + per-response nonce. No unsafe-inline Tailwind ships as an
# external stylesheet, Turbo's injected <style> carries the nonce, and Stimulus
# sets styles via the CSSOM (not governed by CSP).
policy.style_src :self
# Images: Allow self, data URLs, and https for external images
policy.img_src :self, :data, :https
@@ -59,6 +61,13 @@ Rails.application.configure do
policy.report_uri "/api/csp-violation-report"
end
# Per-response random nonce applied to script-src and style-src. The app does
# not page-cache HTML, so a fresh random nonce per response is the most secure
# choice (no reuse across responses). csp_meta_tag (in the layout) and
# importmap-rails read this nonce automatically.
config.content_security_policy_nonce_generator = ->(_request) { SecureRandom.base64(16) }
config.content_security_policy_nonce_directives = %w[script-src style-src]
# Start with CSP in report-only mode for testing
# Set to false after verifying everything works in production
config.content_security_policy_report_only = Rails.env.development?

View File

@@ -4,5 +4,8 @@
# Use this to limit dissemination of sensitive information.
# See the ActiveSupport::ParameterFilter documentation for supported notations and behaviors.
Rails.application.config.filter_parameters += [
:passw, :email, :secret, :token, :_key, :crypt, :salt, :certificate, :otp, :ssn, :cvv, :cvc, :backup
:passw, :email, :secret, :token, :_key, :crypt, :salt, :certificate, :otp, :ssn, :cvv, :cvc, :backup,
# :code partially matches the TOTP/backup `code` param, the OAuth authorization
# `code`, and the PKCE `code_verifier`/`code_challenge` — all sensitive in logs.
:code
]

View File

@@ -1,5 +1,5 @@
# frozen_string_literal: true
module Clinch
VERSION = "0.15.0"
VERSION = "0.16.0"
end

View File

@@ -0,0 +1,7 @@
class AddLastOtpAtToUsers < ActiveRecord::Migration[8.1]
def change
# Unix timestamp of the most recently accepted TOTP timestep, used to reject
# replay of a code within its drift window (passed to ROTP's `after:`).
add_column :users, :last_otp_at, :integer
end
end

3
db/schema.rb generated
View File

@@ -10,7 +10,7 @@
#
# It's strongly recommended that you check this file into your version control system.
ActiveRecord::Schema[8.1].define(version: 2026_06_07_000003) do
ActiveRecord::Schema[8.1].define(version: 2026_06_11_000001) do
create_table "active_storage_attachments", force: :cascade do |t|
t.bigint "blob_id", null: false
t.datetime "created_at", null: false
@@ -233,6 +233,7 @@ ActiveRecord::Schema[8.1].define(version: 2026_06_07_000003) do
t.datetime "created_at", null: false
t.json "custom_claims", default: {}, null: false
t.string "email_address", null: false
t.integer "last_otp_at"
t.datetime "last_sign_in_at"
t.string "name"
t.string "password_digest", null: false

View File

@@ -113,6 +113,42 @@ module Api
assert_equal "Application is inactive", json["error"]
end
test "bearer token returns 401 once user is removed from allowed groups" do
# App restricted to a specific group; user is a member when the key is made.
group = Group.create!(name: "webdav-users")
restricted_app = Application.create!(
name: "Restricted WebDAV",
slug: "restricted-webdav",
app_type: "forward_auth",
domain_pattern: "restricted.example.com",
active: true
)
restricted_app.allowed_groups << group
@user.groups << group
key = @user.api_keys.create!(name: "Restricted Key", application: restricted_app)
token = key.plaintext_token
# Sanity: access works while membership stands.
get "/api/verify", headers: {
"Authorization" => "Bearer #{token}",
"X-Forwarded-Host" => "restricted.example.com"
}
assert_response :ok
# Revoke group membership; the existing key must stop working.
@user.groups.destroy(group)
get "/api/verify", headers: {
"Authorization" => "Bearer #{token}",
"X-Forwarded-Host" => "restricted.example.com"
}
assert_response :unauthorized
json = JSON.parse(response.body)
assert_equal "Access denied: insufficient group membership", json["error"]
end
test "no bearer token falls through to cookie auth" do
# No auth header, no session -> should redirect (cookie flow)
get "/api/verify", headers: {

View File

@@ -242,6 +242,20 @@ module Api
assert_equal "No authentication rule configured for this domain", response.headers["x-auth-reason"]
end
# Fail closed when no host can be determined: emitting identity headers without
# an application would bypass all per-domain group access control.
test "should fail closed and emit no identity headers when host is absent" do
sign_in_as(@user)
# Blank both host sources so forwarded_host is not present.
get "/api/verify", headers: {"X-Forwarded-Host" => "", "Host" => ""}
assert_response 403
assert_equal "No host header present", response.headers["x-auth-reason"]
assert_nil response.headers["X-Remote-User"]
assert_nil response.headers["X-Remote-Groups"]
end
# Security Tests
test "should handle very long domain names" do
long_domain = "a" * 250 + ".example.com"

View File

@@ -34,6 +34,25 @@ class OidcAuthorizationCodeSecurityTest < ActionDispatch::IntegrationTest
# CRITICAL SECURITY TESTS
# ====================
test "consent endpoint rejects cross-site POST without a CSRF token" do
sign_in_as(@user)
# Forgery protection is disabled in the test env by default; enable it so the
# before_action actually runs, mirroring production behaviour.
original = ActionController::Base.allow_forgery_protection
ActionController::Base.allow_forgery_protection = true
begin
# No authenticity_token param: a forged cross-site submission. Because
# :consent is NOT in the verify_authenticity_token skip list, this must be
# rejected before the action can grant any OAuth scopes.
post "/oauth/authorize/consent", params: {approve: "true"}
assert_response :unprocessable_entity
ensure
ActionController::Base.allow_forgery_protection = original
end
end
test "prevents authorization code reuse - sequential attempts" do
# Create consent
OidcUserConsent.create!(

View File

@@ -213,6 +213,50 @@ class OidcRefreshTokenControllerTest < ActionDispatch::IntegrationTest
assert_equal family_id, new_refresh_token.token_family_id
end
test "reusing a revoked refresh token revokes every access token in the family" do
access_token = OidcAccessToken.create!(
application: @application,
user: @user,
scope: "openid profile email"
)
refresh_token = OidcRefreshToken.create!(
application: @application,
user: @user,
oidc_access_token: access_token,
scope: "openid profile email"
)
family_id = refresh_token.token_family_id
old_plaintext = refresh_token.token
# Rotate once: the old refresh token is revoked; a new access + refresh token
# are issued into the same family.
post "/oauth/token", params: {
grant_type: "refresh_token",
refresh_token: old_plaintext,
client_id: @application.client_id,
client_secret: @client_secret
}
assert_response :success
new_refresh = OidcRefreshToken.in_family(family_id).where.not(id: refresh_token.id).first
new_access_token = new_refresh.oidc_access_token
refute new_access_token.reload.revoked?, "rotated-in access token should start active"
# Reuse the OLD (now revoked) refresh token -> rotation-attack detection.
post "/oauth/token", params: {
grant_type: "refresh_token",
refresh_token: old_plaintext,
client_id: @application.client_id,
client_secret: @client_secret
}
assert_response :bad_request
# Both the original and the rotated-in access token must now be revoked, so a
# stolen access token from anywhere in the chain stops working at /userinfo.
assert access_token.reload.revoked?, "original access token should be revoked"
assert new_access_token.reload.revoked?, "rotated-in access token should be revoked"
end
test "userinfo endpoint works with hashed access token" do
access_token = OidcAccessToken.create!(
application: @application,

View File

@@ -19,16 +19,21 @@ class TotpSecurityTest < ActionDispatch::IntegrationTest
# First use of the code should succeed
post totp_verification_path, params: {code: valid_code}
assert_response :redirect
assert_redirected_to root_path
# Sign out
delete session_path
assert_response :redirect
# Note: In the current implementation, TOTP codes CAN be reused within the 60-second time window
# This is standard TOTP behavior. For enhanced security, you could implement used code tracking.
# This test documents the current behavior - codes work within their time window
# Replay the SAME code in a fresh sign-in attempt. Because verify_totp records
# the accepted timestep (ROTP `after:`), the code is now rejected even though
# it is still within its drift window — so we stay on the verification step.
post signin_path, params: {email_address: "totp_replay_test@example.com", password: "password123"}
assert_redirected_to totp_verification_path
post totp_verification_path, params: {code: valid_code}
assert_redirected_to totp_verification_path
assert_equal "Invalid verification code. Please try again.", flash[:alert]
user.sessions.delete_all
user.destroy

View File

@@ -0,0 +1,40 @@
require "test_helper"
class CspTest < ActionDispatch::IntegrationTest
# In the test env content_security_policy_report_only is false, so the enforcing
# Content-Security-Policy header is emitted.
test "signin page sends a nonce-based CSP with no unsafe-inline" do
get signin_path
assert_response :success
csp = response.headers["Content-Security-Policy"]
assert csp.present?, "expected a Content-Security-Policy header"
script_src = directive(csp, "script-src")
style_src = directive(csp, "style-src")
assert_includes script_src, "'nonce-", "script-src must carry a nonce"
assert_includes style_src, "'nonce-", "style-src must carry a nonce"
refute_includes script_src, "'unsafe-inline'", "script-src must not allow unsafe-inline"
refute_includes style_src, "'unsafe-inline'", "style-src must not allow unsafe-inline"
end
test "the inline theme script carries the script-src nonce" do
get signin_path
assert_response :success
header_nonce = response.headers["Content-Security-Policy"][/script-src[^;]*'nonce-([^']+)'/, 1]
assert header_nonce.present?, "expected a nonce in the CSP header"
# The hand-written dark-mode <script> in the layout must use the same nonce,
# otherwise it would be blocked under the enforcing policy.
assert_match(/<script nonce="#{Regexp.escape(header_nonce)}">/, response.body,
"inline theme script must carry the matching CSP nonce")
end
private
def directive(csp, name)
csp.split(";").map(&:strip).find { |d| d.start_with?("#{name} ") } || ""
end
end

View File

@@ -153,6 +153,10 @@ class ForwardAuthIntegrationTest < ActionDispatch::IntegrationTest
# Redirect URL Integration Tests
test "unauthenticated request redirects to signin with parameters" do
# grafana.example.com must be a registered forward-auth app for its URL to be
# honoured as a redirect target (otherwise it would be an open-redirect vector).
grant_everyone_access Application.create!(name: "Grafana", slug: "grafana", app_type: "forward_auth", domain_pattern: "grafana.example.com", active: true)
# Test that unauthenticated requests redirect to signin with rd and rm parameters
get "/api/verify", headers: {
"X-Forwarded-Host" => "grafana.example.com"
@@ -172,7 +176,27 @@ class ForwardAuthIntegrationTest < ActionDispatch::IntegrationTest
assert_includes location, "grafana.example.com"
end
test "spoofed X-Forwarded-Host is not reflected as a redirect target" do
# No forward-auth app exists for evil.com, and no valid rd is supplied. The
# attacker-controlled host must NOT be stored or reflected into the signin URL,
# and base_url must come from CLINCH_HOST (or the safe localhost default in
# test) rather than the request host.
get "/api/verify", headers: {
"X-Forwarded-Host" => "evil.com",
"X-Forwarded-Uri" => "/steal"
}
assert_response 302
assert_match %r{/signin}, response.location
refute_includes response.location, "evil.com"
refute_match(/evil\.com/, session[:return_to_after_authenticating].to_s)
end
test "return URL functionality after authentication" do
# app.example.com must be a registered forward-auth app for its URL to be
# honoured as a redirect target.
grant_everyone_access Application.create!(name: "App FA", slug: "app-fa", app_type: "forward_auth", domain_pattern: "app.example.com", active: true)
# Initial request should set return URL
get "/api/verify", headers: {
"X-Forwarded-Host" => "app.example.com",

View File

@@ -1,6 +1,49 @@
require "test_helper"
class SessionSecurityTest < ActionDispatch::IntegrationTest
# ====================
# ACCOUNT DEACTIVATION TESTS
# ====================
test "TOTP verification rejects a user disabled mid-flow" do
user = User.create!(email_address: "midflow_totp@example.com", password: "password123")
user.enable_totp!
code = ROTP::TOTP.new(user.totp_secret).now
# Phase A: password step stashes the pending 2FA user
post signin_path, params: {email_address: "midflow_totp@example.com", password: "password123"}
assert_redirected_to totp_verification_path
# Admin disables the account while the user is on the 2FA screen
user.update!(status: :disabled)
# Phase B: completing TOTP must NOT create a session
post totp_verification_path, params: {code: code}
assert_redirected_to signin_path
assert_equal 0, user.reload.sessions.count
user.destroy
end
test "an existing session stops authenticating once the user is disabled" do
user = User.create!(email_address: "disabled_session@example.com", password: "password123")
sign_in_as(user)
get root_path
assert_response :success
# Disable bypassing the destroy callback to isolate the request-time lookup
# guard (find_session_by_cookie filtering on active users).
user.update_column(:status, User.statuses[:disabled])
get root_path
assert_response :redirect
assert_match %r{/signin}, response.location
user.sessions.delete_all
user.destroy
end
# ====================
# SESSION TIMEOUT TESTS
# ====================
@@ -199,7 +242,7 @@ class SessionSecurityTest < ActionDispatch::IntegrationTest
slug: "logout-test-app",
app_type: "oidc",
redirect_uris: ["http://localhost:4000/callback"].to_json,
backchannel_logout_uri: "http://localhost:4000/logout",
backchannel_logout_uri: "https://rp.example.com/backchannel-logout",
active: true
)

View File

@@ -0,0 +1,38 @@
require "test_helper"
class PrivateAddressCheckTest < ActiveSupport::TestCase
# internal_host? — DNS-free checks on IP literals and known hostnames
test "flags loopback, private, and link-local IP literals as internal" do
%w[
127.0.0.1
10.0.0.1
172.16.5.5
192.168.1.1
169.254.169.254
0.0.0.0
::1
].each do |host|
assert PrivateAddressCheck.internal_host?(host), "expected #{host} to be internal"
end
end
test "flags localhost-style hostnames as internal" do
assert PrivateAddressCheck.internal_host?("localhost")
assert PrivateAddressCheck.internal_host?("foo.localhost")
assert PrivateAddressCheck.internal_host?("metadata.google.internal")
assert PrivateAddressCheck.internal_host?("")
end
test "does not flag public IP literals as internal" do
refute PrivateAddressCheck.internal_host?("8.8.8.8")
refute PrivateAddressCheck.internal_host?("1.1.1.1")
end
# resolves_to_internal? on IP literals (no DNS needed) exercises the same
# address classification used after resolution.
test "resolves_to_internal? classifies IP literals" do
assert PrivateAddressCheck.resolves_to_internal?("169.254.169.254")
assert PrivateAddressCheck.resolves_to_internal?("127.0.0.1")
refute PrivateAddressCheck.resolves_to_internal?("8.8.8.8")
end
end

View File

@@ -54,6 +54,15 @@ class SecurityMailerTest < ActionMailer::TestCase
assert_bodies_contain email, "Old MacBook"
end
test "suspicious_passkey_used warns about a blocked clone sign-in" do
email = SecurityMailer.suspicious_passkey_used(@user, nickname: "Yubikey-5", **CONTEXT)
assert_equal [@user.email_address], email.to
assert_match(/blocked/i, email.subject)
assert_bodies_contain email, "Yubikey-5"
assert_bodies_match email, /clon/i
end
test "api_key_created includes the key name" do
email = SecurityMailer.api_key_created(@user, name: "CI bot", **CONTEXT)

View File

@@ -56,4 +56,29 @@ class ApplicationTest < ActiveSupport::TestCase
tempfile&.close
tempfile&.unlink
end
test "rejects backchannel_logout_uri pointing at internal addresses (SSRF guard)" do
app = applications(:kavita_app)
internal_uris = [
"http://127.0.0.1/logout",
"http://localhost/logout",
"https://169.254.169.254/latest/meta-data/",
"http://10.0.0.5/logout",
"http://192.168.1.10/logout"
]
internal_uris.each do |uri|
app.backchannel_logout_uri = uri
refute app.valid?, "expected #{uri} to be rejected"
assert_includes app.errors[:backchannel_logout_uri].join, "private, loopback, or link-local"
end
end
test "allows backchannel_logout_uri pointing at a public host" do
app = applications(:kavita_app)
app.backchannel_logout_uri = "https://relying-party.example.com/backchannel-logout"
assert app.valid?, app.errors.full_messages.to_sentence
end
end

View File

@@ -1,6 +1,27 @@
require "test_helper"
class UserTest < ActiveSupport::TestCase
test "disabling a user destroys their active sessions" do
user = User.create!(email_address: "disable_sessions@example.com", password: "password123")
user.sessions.create!
user.sessions.create!
assert_equal 2, user.sessions.count
user.update!(status: :disabled)
assert_equal 0, user.reload.sessions.count
end
test "reactivating or other updates do not destroy sessions" do
user = User.create!(email_address: "keep_sessions@example.com", password: "password123")
user.sessions.create!
# An update that does not change status must leave sessions intact.
user.update!(username: "keepsessions")
assert_equal 1, user.reload.sessions.count
end
test "downcases and strips email_address" do
user = User.new(email_address: " DOWNCASED@EXAMPLE.COM ")
assert_equal("downcased@example.com", user.email_address)

View File

@@ -0,0 +1,29 @@
require "test_helper"
class WebauthnCredentialTest < ActiveSupport::TestCase
# suspicious_sign_count?(new_sign_count) — clone detection per WebAuthn §6.1.1.
# Build an in-memory credential with a given stored sign_count; no persistence
# needed since the method only reads self.sign_count.
def credential(stored:)
WebauthnCredential.new(sign_count: stored)
end
test "does not flag when the authenticator reports no counter (synced passkeys)" do
# Both 0 -> authenticator doesn't implement a counter; must NOT be suspicious.
refute credential(stored: 0).suspicious_sign_count?(0)
# Stored 0, first real use.
refute credential(stored: 0).suspicious_sign_count?(5)
# Stored non-zero but authenticator now reports 0 -> no counter, not a clone.
refute credential(stored: 5).suspicious_sign_count?(0)
end
test "does not flag a normal increasing counter" do
refute credential(stored: 5).suspicious_sign_count?(6)
refute credential(stored: 1).suspicious_sign_count?(1000)
end
test "flags a non-advancing counter as a possible clone" do
assert credential(stored: 5).suspicious_sign_count?(5), "equal count is suspicious"
assert credential(stored: 5).suspicious_sign_count?(3), "decreasing count is suspicious"
end
end