What 250,000+ protected sites and a measured test of WP Ghost 9.1.08 show
Last updated: October 28, 2026 · Network data: last 30 days (GDPR-compliant retention) · Lab test: WP Ghost 9.1.08
WP Ghost blocks more than 100 million threat attempts every month across the 250,000+ WordPress sites running it, including more than 10 million brute force attempts. Those two figures come from cloud reporting on sites that have the Security Threats Log turned on.
They’re the floor, not the ceiling. Path Security and the server-level firewall rules stop most attacks before the logging code runs, so those blocks never reach the count. The “How we measure” section below explains where the gap comes from.
To measure what telemetry can’t see, we also tested WP Ghost 9.1.08 on a WooCommerce test shop in October 2026 and ran the same attacks with the plugin on and off. The lab results sit next to the network numbers, and each one is labeled with its source.
The headline numbers
From the WP Ghost network (last 30 days)
- 100M+ threat attempts blocked by WP Ghost’s Application Firewall (7G, 8G and 2026 Ghost Firewall levels) and security rules across the network
- 10M+ brute force attempts stopped by WP Ghost’s login protection across the network
- 250,000+ WordPress sites actively protected by WP Ghost
From the WP Ghost 9.1.08 lab test (October 2026)
- 24 of 24 SQL injection payloads blocked before WordPress loaded
- 0 false positives on 38 legitimate requests, including “how to delete a user” and “order by relevance”
- 0.0 seconds of PHP CPU used to absorb 10,000 attacks, against 522 seconds with the plugin off
- 100,000 attacks blocked in 36 seconds, all of them before PHP started
- 10 of 10 AI training crawlers blocked, while Googlebot, Bingbot and human visitors passed
- 4 ms added to a WooCommerce product page (68 ms median with Ghost Mode, 64 ms without the plugin)
How we measure
The data sources
Source 1: WP Ghost cloud telemetry
Sites with the Security Threats Log and cloud reporting turned on send two data points per day to the WP Ghost Dashboard: the date and the total count of blocked threats. No IP addresses, URLs, request payloads or visitor data are sent. The WP Ghost GDPR Compliance page documents the full list.
We keep this data for 30 days and then delete it, so the network numbers are a rolling 30-day snapshot. We refresh them monthly.
Source 2: The WP Ghost 9.1.08 lab test
The test shop ran WordPress 7.1.3 and WooCommerce 11.2.1 with the Storefront theme and no page cache. The server was a Hetzner VPS with 2 vCPU and 4 GB of RAM, running Ubuntu 26.04 and PHP 8.5.4 on php-fpm with 8 workers.
WP Ghost 9.1.08 ran in Ghost Mode with every feature on and its rules in the Apache .htaccess file. Traffic came over loopback, so the millisecond figures are server and PHP cost only.
| Test setting | Value |
|---|---|
| WordPress | 7.1.3 |
| WooCommerce | 11.2.1, Storefront theme |
| Page cache | None |
| Server | Hetzner VPS, 2 vCPU, 4 GB RAM |
| Stack | Ubuntu 26.04, PHP 8.5.4, php-fpm with 8 workers |
| WP Ghost | 9.1.08 (build 7561aa9), Ghost Mode, all features on |
| Firewall rules | Apache .htaccess |
How the lab test scores an attack
Each attack counts as stopped only when it fails to reach its goal. A 403 status code alone doesn’t decide it.
| Attack type | Counted as stopped when |
|---|---|
| Injection payload | The firewall returns 403 before WordPress loads |
| User enumeration | The admin username can’t be read |
| Open redirect | The browser isn’t sent to another site |
| File probe | The file isn’t served |
Some protections defuse an attack and still answer 200. The user enumeration filter returns an empty list, for example. Counting status codes alone would undercount those.
What we can measure
- Threats caught by the Application Firewall (7G, 8G or 2026 Ghost Firewall level) at the PHP level
- Brute force attempts on the login, password reset, registration and comment forms
- Security rule triggers: SQL injection and XSS patterns, malicious user agents, suspicious request parameters
- IP addresses caught by Automated IP Blocking before the server-level ban
What we can’t measure
Path Security blocks never reach the network count. When WP Ghost changes /wp-login.php, /wp-admin, /wp-content, /wp-includes, /uploads, /plugins and /themes, requests to the default paths get a 404 from the web server. WordPress doesn’t load, and the logging code never runs.
The same goes for firewall rules in .htaccess or the Nginx config. Apache or Nginx rejects the request before PHP starts, so nothing is written to the log.
Automated IP Blocking logs the first attack from an IP. After the ban, the next thousand requests from that IP are refused at the server level and never counted.
So the 100M+ figure covers attacks that got past Path Security and the server rules and were then caught at the application layer. The lab test is how we measure the earlier layers: on the test shop, the 10,000-attack run used 0.0 seconds of PHP CPU because the server refused every request before PHP started.
Lab results: file and probe blocking
The test used files that exist on the shop, so a block means a real file was refused, not that it was missing.
| Target | Result |
|---|---|
| Sensitive files on default paths (readme.html, license.txt, wp-config-sample.php, wp-includes/version.php, install.php and 4 more) | 9 of 9 refused (404 or 403) |
| Scanner filenames and probe paths (shell.php, wso.php, alfa.php, PHP in /wp-content/uploads, /.env, /.git/config) | 6 of 6 refused |
| Shop pages (home, shop, product, cart, theme CSS and scripts) | All loaded |
Renaming the asset paths themselves, so the CSS and JS URLs change too, turns on through the one-click frontend test in the WP Ghost admin. That step wasn’t run from the command line, so the table covers file and probe blocking only.
Lab results: 2026 Ghost Firewall
The corpus held 85 attack payloads across nine classes, plus 38 legitimate requests. Settings for this level are in the 2026 Ghost Firewall guide.
| Attack class | Blocked |
|---|---|
| SQL injection (classic, time-based, boolean) | 24 of 24 |
| Cross-site scripting | 10 of 10 |
| Path traversal and local file inclusion, including double-encoded | 9 of 9 |
| PHP object injection, including base64-wrapped | 3 of 3 |
| Secret and config file probes | 13 of 13 |
| Command injection | 6 of 8 |
| Server-side request forgery | 4 of 5 |
| False positives on 38 legitimate requests | 0 |
The two command injection payloads that passed use shell command substitution (backticks and dollar-paren). WordPress never hands request data to a shell, and the rule stays off so it doesn’t break legitimate content or cron jobs.
The one SSRF payload that passed is a protocol-relative open redirect. The redirect check below stops it.
Three related protections were scored by whether the attack worked:
| Attack | Result with WP Ghost on |
|---|---|
| REST API user enumeration | Empty list returned (the admin username shows with the plugin off) |
| Logout with redirect_to=//evil.com | Sent to the site’s own URL |
| fwrite(fopen(‘shell.php’,’w’),…) payload | 403 |
An internal redirect_to=/my-account/ still worked as normal.
Lab results: bot and brute force blocking
Automated IP Blocking stopped a login attacker at the fifth failed attempt and then locked that IP out of the whole site.
The AI crawler block stopped 10 of 10 training and scraping bots: GPTBot, ClaudeBot, CCBot, Bytespider, PerplexityBot, Meta-externalagent, Google-Extended, Amazonbot, Applebot-Extended and cohere-ai. Googlebot, Bingbot and human visitors passed.
Throttle AI Training Crawler Bots was tested with 10 training crawlers hitting the shop at once:
| Under 10 concurrent AI crawlers | Throttle off | Throttle on |
|---|---|---|
| PHP CPU | 26.5 s | 12.9 s |
| Full page builds | 692 | 190 |
| php-fpm workers busy | 7 of 8 | 3 of 8 |
| Page load for a real visitor | 250 ms | 66 ms |
Search engines, human visitors and live AI user requests are never throttled.
Lab results: spoofing tests
The same XSS and SQL injection attack got a 403 under four user agents: a plain browser, a fake Googlebot, a fake GPTBot and sqlmap.
Setting X-Forwarded-For or X-Real-IP to a whitelisted address still got a 403. WP Ghost trusts only the real connection address.
A real Googlebot user agent on a clean request got a 200.
Lab results: server load under 10,000 attacks
| 10,000 attacks | PHP CPU | php-fpm workers | Wall time | Throughput |
|---|---|---|---|---|
| WP Ghost on | 0.0 s | 0 of 8 | 4 s | 2,810 req/s |
| WP Ghost off (WooCommerce renders each request) | 522 s | 8 of 8 | 302 s | 33 req/s |
With the plugin off, WooCommerce built a full page for every attack and kept all 8 workers busy for five minutes. With WP Ghost on, the run finished in 4 seconds and PHP never started.
A second run sent 100,000 attacks. WP Ghost blocked all of them in 36 seconds with zero PHP CPU.
Lab results: page speed and the shop
A WooCommerce product page loaded in 68 ms (median of 40 loads) with Ghost Mode and every feature on. With the plugin deactivated, it loaded in 64 ms.
The 4 ms difference comes from the Ghost Mode URL rewriting. The firewall on its own changed throughput by under 1% across 2,000 clean requests.
Checkout kept working with WP Ghost active. A cart-to-order test with Cash on Delivery created order #12, status processing, for $58.00. The cart, checkout and My Account pages all loaded.
WP Super Cache installed and activated cleanly on the same shop, and the firewall kept blocking with it on.
Limits of the lab test
The test ran on one WooCommerce shop on one server. A site with a different plugin stack or host can get different numbers.
Loopback removes network latency. A real visitor adds the same latency to every row, so the on/off ratios hold.
The 302-second “plugin off” time for 10,000 attacks was measured. The 100,000-attack run was measured with WP Ghost on only.
An earlier claim that a whitelisted IP skips the firewall is withdrawn. In testing, an attack from a whitelisted IP still got a 403, and the decision path couldn’t be traced, so we don’t claim the exemption.
Why the network totals look the way they do
The 100M+ monthly figure counts attacks that reached the Application Firewall. Path Security and the server-level rules had already rejected the default-path probes and the plain malicious requests.
Other WordPress security vendors publish the same kind of telemetry. Wordfence reported blocking 10.4 billion WAF attacks and 18.2 billion brute force attacks across its network in Q2 2026 (Wordfence, September 29, 2026). Network totals like these show how much automated traffic hits WordPress, not how many sites were breached.
What this data doesn’t prove
WP Ghost prevents attacks. It doesn’t:
- Scan your site for existing malware (use Wordfence, Sucuri or MalCare)
- Clean an infected site (use Sucuri’s cleanup service or a specialist)
- Back up your content (use UpdraftPlus, BlogVault or your host’s backups)
- Patch vulnerable plugin code (keep your plugins updated)
Results on a single site depend on its public profile and on the security level picked (Lite Mode, Safe Mode or Ghost Mode). They also depend on the plugin and theme stack, and on whether the site was already compromised before WP Ghost went on.
How this data is collected
WP Ghost operates under EU data protection rules (GDPR). MINBO QRE SRL, the company behind WP Ghost, is registered in Romania.
Network telemetry is opt-in and minimal:
- It runs only when a site turns on the Security Threats Log with cloud reporting
- Each site sends two data points per day: the date and the total threat count
- No IP addresses, URLs, request payloads, user data or personal information are collected
- The data is deleted automatically after 30 days
- Cloud reporting can be turned off at any time in the WP Ghost settings
The lab traffic was generated for the test on a test shop, so no visitor data was involved.
Frequently asked questions
Does WP Ghost actually stop attacks?
In the October 2026 lab test, WP Ghost 9.1.08 blocked 24 of 24 SQL injection payloads and 10 of 10 XSS payloads, with zero false positives on 38 legitimate requests. Across the network it blocks more than 100 million threat attempts a month.
Is the 100M+ monthly threats number real?
Yes. It’s a 30-day total from sites with the Security Threats Log and cloud reporting on, counted at the Application Firewall layer. Blocks by Path Security and server-level rules happen earlier and aren’t included, so the real number is higher.
What data does WP Ghost collect?
Only the date and the total count of blocked threats, once per day, and only when cloud reporting is on. Details are on the WP Ghost GDPR Compliance page.
Why does WP Ghost count fewer threats than it blocks?
Path Security and the .htaccess firewall rules reject requests before PHP starts, so the logging code never sees them. In the lab, 10,000 attacks used 0.0 seconds of PHP CPU for that reason.
Does Path Security alone stop anything?
On the test shop it refused 9 of 9 sensitive files and 6 of 6 scanner probes on the default paths. It works best with the 2026 Ghost Firewall and brute force protection turned on as well.
Does WP Ghost slow down WordPress?
In the lab, a WooCommerce product page went from 64 ms to 68 ms median with Ghost Mode and every feature on. The firewall alone changed throughput by under 1%.
Can a fake Googlebot get past the firewall?
No. An attack sent as a fake Googlebot or a fake GPTBot still got a 403, and so did one with a spoofed X-Forwarded-For header.
How does WP Ghost compare to a malware scanner?
A malware scanner like Wordfence or Sucuri finds and cleans files already placed on a site, and WP Ghost blocks the attacks that place them. WP Ghost is compatible with Wordfence, Sucuri, Solid Security, Cloudflare, BBQ Firewall and SiteGround Security.
Why isn’t there a 12-month trend chart?
Cloud reporting data is deleted after 30 days, in line with GDPR data minimization. For year-long WordPress attack data, see our WordPress Security Statistics 2025-2026 report.
Can I see my own site’s data?
Yes. WP Ghost Premium has the Security Threats Log (30 days of events with search, CSV export and filters) and the GEO Threat Map. The free version shows the latest events on the dashboard.
Who is behind these numbers?
WP Ghost is built by MINBO QRE SRL, a software company in Romania. The network figures come from WP Ghost cloud telemetry, and the lab figures come from our test of version 9.1.08.
Industry context
External research on WordPress attacks:
- 11,334 new WordPress vulnerabilities were recorded in 2025, up 42% on 2024 (Patchstack, February 2026)
- 91% of those vulnerabilities were in plugins and 9% in themes (Patchstack 2026)
- 46% had no fix from the developer at the time of public disclosure (Patchstack 2026)
- Heavily exploited vulnerabilities reached mass exploitation in a weighted median of 5 hours (Patchstack 2026)
- In Patchstack’s test against known exploited WordPress vulnerabilities, traditional hosting defenses blocked 12% of attacks (Patchstack 2026)
- Wordfence blocked 10.4 billion WAF attacks and 18.2 billion brute force attacks in Q2 2026 and counted 573,000 infected sites (Wordfence, September 2026)
- 2,073 new WordPress vulnerabilities were published in Q2 2026, down 24.3% from Q1 (Wordfence)
The full industry analysis is in WordPress Security Statistics 2025-2026.
How this page is maintained
| Item | Value |
|---|---|
| Network data window | Last 30 days, GDPR retention |
| Network refresh | Monthly |
| Lab test | WP Ghost 9.1.08, October 2026 |
| Data owner | WP Ghost Team, MINBO QRE SRL |
| Methodology changes | Logged in the page’s revision history |
When the Security Threats Log adds threat-type categories, the breakdown goes on this page with the date the method changed. Version changes are listed in the WP Ghost changelog.