WordPress Security Checklist for 2026: What We Changed After Two Breaches

Todd Zeigler Avatar

|

Estimated Reading Time: 7 minutes

Summary: Attacks on WordPress sites got faster and more automated over the last year, and we suspect AI is the reason. Below is what happened in two incidents we handled this summer, followed by the WordPress security checklist we now work from.

When a big organization has a bad week online, it makes the news. Microsoft’s email goes down, or the FAA grounds flights, and the first question everyone asks is whether it was a hack. Often it turns out to be a software bug. Sometimes it is an attack. These are organizations with security teams larger than most of our clients’ entire staff, and they still get caught out.

You can’t do anything about their systems. Your own website is the part you control, and for most of the time we have worked with WordPress, keeping it up to date was enough to keep it out of trouble.

Over the last year that stopped being true. Updates are still the foundation, but attacks now arrive within hours of a vulnerability being disclosed, and in volumes we haven’t seen before. We got our own wake-up call over the summer, when we had to deal with two security incidents. In this post, I explain why attacks are getting more frequent and outline the steps we now take to guard against them.

Why Attacks Got Faster

For most of WordPress’s history, a vulnerability disclosure gave you a window. A patch would come out, and it would take days or weeks for a working exploit to show up. If you patched within a day or two you were generally fine.

That window is closing. In July, a WordPress security release came out on the 17th. The first attempts against our sites started at 21:29 UTC the same day, and over the following week we logged roughly 37,000 attempts against about 30 hostnames.

We suspect AI tooling is doing a lot of this work. Reading a patch, figuring out what it fixes, and writing code to exploit the unpatched version is the kind of task current AI models are good at. Same-day exploits are not new, but they used to be rare.

Two Incidents from 2026

Incident 1: Attacked the Same Day the Patch Shipped

On July 17, WordPress released version 7.0.2 to fix two vulnerabilities in WordPress core, CVE-2026-60137 and CVE-2026-63030, known together as wp2shell. Chained together, they let an attacker with no account read from and write to the WordPress database. We had the patch applied across our client sites within about 90 minutes of the advisory. The attack was already underway.

The results split by environment. On a development server, the exploit succeeded and a cryptocurrency miner ran for a week before an AWS billing alert caught it. Several dev sites were affected, all copies of production. On the live sites, one unauthorized admin account was created, which we removed. Every attempt to install a backdoor failed, because the production file systems are read-only and the web server couldn’t write code to disk. No malware landed on production and we found no evidence of a data leak.

Our response time was good and it didn’t matter. What protected production was a hosting decision made years earlier. What exposed the dev servers was that nobody had applied the same standard to them.

Incident 2: A Stolen Admin Password

The second incident had nothing to do with a WordPress vulnerability. An administrator’s computer was infected with a trojan that captured saved passwords, including the one for the site. The attacker logged in through the normal login form with valid credentials.

A good lock doesn’t help much if someone copies the key. Two-factor authentication would have stopped the login outright, and an alert on new admin accounts would have caught the intrusion in under 35 seconds instead of days later. Neither was in place on that site.

The Checklist

We don’t do all of this for every site. What we recommend depends on how much damage a breach would do, what the hosting platform allows, and what the client wants to spend. The first group is cheap enough that we now push for it on nearly every site. The second group is a conversation about risk and budget.

None of this replaces keeping WordPress, your theme, and your plugins current. The basics in our 2019 WordPress security tips still apply.

StepSoftware costWho it’s for
Two-factor authentication on all admin accountsFreeEvery site
Disable XML-RPCFreeEvery site
Block username enumerationFreeEvery site
Change the login URLFreeEvery site, after the three above
Password-protect development sitesFreeEvery site with a dev environment
Audit hosting account accessFreeEvery site
Read-only production file systemDepends on hostSites where a compromise would be costly
Remove plugins that run code from the databaseFreeSites where a compromise would be costly
Web application firewall with virtual patching$20 to $25/month per siteSites where a compromise would be costly
Alerts on plugin changes and new adminsFreeSites where a compromise would be costly
Core update monitoringFreeSites on hosting that blocks auto-updates
Access layer in front of wp-adminFree tier availableSites with a small number of admins
Authenticate outgoing emailFree to $25/monthSites that email login codes

Worth Doing on Almost Any Site

Enable Two-Factor Authentication on Every Admin Account

If you do one thing on this list, do this. With 2FA in place, a stolen password is not enough to get into your site. Use an authenticator app where you can. While you are at it, replace shared logins like webadmin@ with named accounts, and reset every admin password once 2FA is on.

Disable XML-RPC

XML-RPC is an old remote publishing interface that almost nobody uses anymore. Unfortunately, it accepts a username and password directly and skips your login form, which means it skips 2FA. It takes minutes to turn off. Check first whether the site uses Jetpack, which depends on it.

Block Username Enumeration

By default, WordPress exposes author names in several places, and on many sites the author name is also the login username. An attacker with a valid username is halfway to a login. We use a custom plugin that blocks the common ways WordPress gives usernames away.

Change the Default Login URL

Moving your login page off /wp-login.php kills the undirected brute force traffic that hits every WordPress site on the Internet. Note that this only holds up once XML-RPC is disabled and usernames are hidden. On its own it is a speed bump.

Put Development Sites Behind a Password

Development and staging sites are copies of production, including the database. That means they contain your real user table and your real password hashes. We found dev sites that were publicly reachable with exactly that in them. A password in front of the dev site closes the gap.

Audit Who Has Access to Your Hosting Account

Hosting access accumulates. When we reviewed one hosting organization we found 24 people with access, 4 of them at outside email addresses. Most had a legitimate reason at some point and nobody ever removed them. Review the list at least once a year and cut everyone down to the access their role needs.

Worth Considering Based on Risk and Budget

Make the Production File System Read-Only

This is the change that mattered most in July. On a read-only host, the web server cannot write program code to disk. An attacker can get in, but they can’t leave anything behind. Platforms like Pantheon and Upsun do this by default. The tradeoff is that you lose one-click plugin installs and WordPress’s emergency auto-updates.

Remove Plugins That Run Code from the Database

There is one exception to the protection read-only hosting gives you. Plugins like WPCode store PHP snippets in the database and run them on every page load. An attacker with database access, which is what the July exploit provided, can plant code there without touching the file system. Move the snippets into the theme or a proper plugin and remove the snippet plugin.

Add a Web Application Firewall with Virtual Patching

A web application firewall (WAF) sits in front of your site and blocks requests that match known attack patterns. The feature that matters is virtual patching. When a vulnerability is disclosed, WAF vendors push a rule that blocks the exploit within hours, usually before you have patched the site yourself. It does not replace updating, but it covers the window described above. Expect $20 to $25 per month per site.

Set Up Alerts for Plugin Uploads, Deactivations, and New Admin Accounts

Both intrusions we handled this summer were over in under 35 seconds, and nobody noticed either one until well after the fact. An attacker who gets in will typically upload a plugin, deactivate a security plugin, or create an admin account. All three can be sent to Slack or email in real time.

Keep WordPress Core Current and Monitor It

This sounds obvious. The catch is that read-only hosting blocks the emergency auto-updates WordPress pushes for security releases, so somebody has to apply them by hand. In one portfolio we found 6 sites running core versions between 18 months and 4 years old. Set up monitoring so you know within a day when a core release ships.

Put an Access Layer in Front of wp-admin

Tools like Cloudflare Zero Trust Access require a separate login, tied to your company email, before anyone can reach the WordPress admin. On our own site this cut login attempts from thousands per month to 34. It works best on sites with a small number of admins and is a poor fit for membership sites.

Authenticate Your Site’s Outgoing Email

WordPress relies on email for password resets, security alerts, and sometimes login codes. On one site we found login codes were failing silently because the site’s mail was not properly authenticated. Set up DKIM and DMARC for the sending domain and turn on mail logging so you can see failures.

Frequently Asked Questions

Is WordPress secure in 2026?

WordPress core is well maintained, and WordPress.org added automated security reviews for plugin releases this year. A site that is kept current and hardened along the lines above can be secured to a high standard. The July incident shows that patching alone will sometimes not be fast enough, which is why this checklist exists.

Do I need a WAF if my host already provides security?

It depends on what the host provides. Read-only file systems, network firewalls, and DDoS protection are common, but virtual patching often isn’t included. Ask your host whether they push rules for newly disclosed WordPress vulnerabilities. If the answer is no, a WAF fills that gap.

How often should I audit my WordPress site’s security?

We run a scan and a configuration review quarterly on the sites where a breach would do the most damage, and we recommend the same for any site with something to lose. Pair it with a short written plan for what to do when something is found.

What Changed for Us

Two assumptions now sit under everything we do with WordPress:

  1. You will sometimes be attacked before you can patch. Build the site so that a successful exploit can’t leave anything behind.
  2. An admin password will eventually leak. Make sure a password alone isn’t enough to get in, and make sure you find out within minutes when someone does.

How far down the checklist a given site should go is still a judgment call. We are also building tools to shorten the time between a vulnerability being disclosed and a tested fix going live, including an AI-assisted process for checking which of our sites are exposed. That work is in progress.