tlloancy Behavioral Ultimate Finisher

By Thomas Lloancy

0+ active installationsVersion 1.2.0

Invisible behavioral spam filter for comments and forms. No CAPTCHA, no cookies, no external service. Pairs well with your anti-spam plugin.

Description

tlloancy Behavioral Ultimate Finisher looks at how a form was filled, not only at what was sent. It scores each submission from a few behavioral signals and lets you decide what happens to the ones that look automated. No CAPTCHA, no cookies, no external service.

A companion for your anti-spam plugin

Honeypots, nonces and form timers are good against simple bots, but a bot that runs JavaScript can satisfy them. Behavioral Ultimate Finisher adds the missing angle: a bot that loads the page, fills the form in a fraction of a second and never touches the mouse or the keyboard gives itself away by how it behaved.

It uses standard WordPress hooks and its own field names, so it is designed to sit next to plugins such as La Sentinelle antispam, Antispam Bee or Akismet. Each plugin runs its own check, and a submission is stopped if either one flags it. This plugin is independent and is not affiliated with those projects.

What it catches

  • Bots that run JavaScript but do not behave like a person: submitting within a fraction of a second, no mouse, keyboard, touch or scroll, scripted keystrokes, headless-browser windows.
  • Raw bots that post straight to the form without a browser, through the honeypot or through the missing token and missing counters added together.

What it does not catch

Bots that drive a real browser with genuine input events, and human spammers, can look like a normal visitor. Keep a content-based anti-spam plugin for those.

Signals (each with an editable score)

  • Honeypot field filled in
  • No valid token (direct POST without a browser, forged or expired token)
  • No behavior counters at all (the script did not run, or the form was posted directly)
  • Form submitted faster than a minimum fill time
  • No human interaction before the submission (mouse, keyboard, touch, scroll, focus)
  • Browser reports automation (navigator.webdriver)
  • Browser window reports a size of 0 x 0, typical of headless browsers
  • Many keystrokes or clicks fired by a script instead of the browser (simulated typing)
  • The same token reused by many submissions (a bot replaying one token)

Scores add up. A submission is flagged when its total reaches your threshold, so one weak signal never blocks a real visitor on its own. The log shows the real sum of the scores, even above 100.

Signal radar

The settings page draws the weight of each signal on a radar chart, with the flag threshold as a dashed ring. It updates as you edit the scores. The helmet in the middle shows how strict your settings are: it glows teal when the filter is calm and turns amber, then red, as it gets stricter, while the crest on its forehead lowers its arms from a Y to a cross.

Three modes

  • Reject (default): every flagged submission is rejected.
  • Mark as spam: flagged comments go to the spam queue, other forms reject the submission.
  • Observe: scores are logged, nothing is blocked. Useful to check the log and calibrate your thresholds on your own traffic.

Protected forms

  • WordPress comments (including WooCommerce reviews)
  • User registration (WordPress and WooCommerce)
  • Contact Form 7
  • Login
  • Lost password (a separate page: ticking Login does not cover it)

For developers

$result = tlbuf_evaluate( 'my-plugin' ); returns `score` (0-100), `total` (real sum), `signals` and `flagged` for the current submission. Use the `tlbuf_behavior_score` filter to adjust the score.

Privacy

No cookies, no IP address stored, no data sent to any third-party service. The log only keeps the date, form type, score, decision, triggered signals and a short line of event counters (numbers only), and is purged automatically. The score and signals of an accepted comment or new user are also kept as comment and user metadata, so you can see why a message got through.

The content log is off by default. When you enable it, the text of flagged or rejected comments (name, URL and message, never the e-mail address or IP) and Contact Form 7 messages is also kept, cut short, for the retention period you set. This is personal data: mention it in your privacy policy. A suggested text is added to Settings > Privacy. Login, lost password and registration are never kept, nor any field that looks like a password, card or token.

Locked out?

Add define( 'TLBUF_DISABLED', true ); to wp-config.php to turn off every check and the tracker. Remove it once you have fixed the settings.

Installation

  1. Upload the plugin folder to /wp-content/plugins/, or install it from the Plugins screen.
  2. Activate it.
  3. Open Settings > Behavioral Ultimate Finisher. The plugin starts in Reject mode, so flagged submissions are blocked right away.
  4. Optional: switch to “Observe” or “Mark as spam” if you want to review the log first.

FAQ

Does it replace my anti-spam plugin?

No, it is meant to work next to it. A content-based plugin judges what was sent, Behavioral Ultimate Finisher judges how the form was filled. You can switch to observe mode at any time to see what it would flag on your own traffic.

Will it block real visitors?

Not on a single signal. A visitor whose script ran but whose token request was blocked (REST API closed, for example) only gets the “no valid token” score and passes. A visitor with JavaScript disabled has neither token nor counters, so by default the two scores add up to the threshold and the submission is rejected. If your audience includes people without JavaScript, lower the “no behavior counters” score, or use observe mode and check the log first.

Is the login form protected?

Yes on new installs, together with the lost password form. Sites that update keep their current login setting, and the lost password form stays off for them until they tick it. Password managers that fill and submit a login form within a second or two can look like a bot, so check the log, and keep the TLBUF_DISABLED switch in mind in case an administrator is locked out.

Does it work with page caching?

Yes. The token is requested from a REST endpoint after the page loads, so cached pages are fine. If a security plugin blocks the REST API for visitors, no token is issued and only the honeypot and the other signals apply. Plugins that delay JavaScript until the first interaction should exclude tlbuf-tracker.js.

Does it stop every bot?

No. Bots that drive a real browser with genuine input events, and human spammers, can pass behavioral checks. Use it as a first filter next to a content-based anti-spam plugin. Tokens are valid for 24 hours and are not single-use.

Does it use an external service or cookies?

No.

Screenshots

Changelog

1.2.0

  • New: lost password form protection (separate setting).
  • New: login and lost password are protected by default on new installs. Existing sites keep their settings.
  • New: the log shows the real total score instead of capping it at 100.
  • New: “token replayed” signal, for a token reused by more than 5 submissions.
  • New: score and signals saved on comments and users, with a Behavior column on the Comments and Users screens.
  • New: optional content log of flagged or rejected comments and Contact Form 7 messages, off by default, with suggested privacy policy text.
  • New: Site Health test for the token endpoint.
  • New: TLBUF_DISABLED constant to switch the plugin off from wp-config.php.
  • Changed: the log label “Passed” is now “Passed this filter”, since a registration is logged before WordPress creates the account.
  • Updated the French translation.

1.1.0

  • New: a “no behavior counters” score. Added to a missing token, it reaches the default threshold, so a direct post is rejected. Visitors without JavaScript are rejected too: lower this score if they must pass.
  • New: a signal radar on the settings page. The helmet shows how strict the filter is.

1.0.0

  • First release.