The patched version was publicly exploited the day after release, which door should the official website close on patch day?
A patch release with a vulnerability fix was publicly exploited the next day, and the official website should first shut down the 'remotely triggerable entry point' on patch day, not all features at once.For example, with this round of issues with WordPress: two unauthenticated SQL injection vulnerabilities CVE-2026-63030 and CVE-2026-60137 affect WordPress 6.9.0-6.9.4 and 7.0.0-7.0.1, three emergency patch versions were released on July 17, 2026 (Friday), and versions after 6.8.6, 6.9.5, and 7.0.2 have been fixed;Public verification code starts spreading at around 01:00 in the early morning of the next day (Saturday). From 'patch available' to 'script available', it only took one day, this is the reality of patch day.
The day most easily overlooked in the timeline
Site administrators usually watch two time points: the time the vulnerability is disclosed and the time the patch is released. The third point that truly determines the risk curve is when the publicly available exploit code appears.
When the fixed version is released, the version number, change content, and affected range are all clear, and the attacker only needs to restore the difference to trigger the condition. After the public exploitation code appears, the technical threshold for exploitation drops from 'being able to find vulnerabilities' to 'being able to run scripts', and un-upgraded sites begin to be scanned and hit in batches. In the above case, this interval is less than a day, and scanning and batch exploitation usually spread faster than manual verification.
Therefore, 'we haven't received an attack yet' cannot be used as an excuse to delay the upgrade: the probing behavior of batch exploitation often appears before business anomalies, and by the time the anomaly is visible, the entry has already been tested.
What should be turned off first when the patch has not been applied?
The order is arranged according to whether it can be triggered by unauthenticated requests.
| Priority | Convergence Action | Why is it effective? | Side effects |
|---|---|---|---|
| First related | Unauthenticated public write interfaces, batch query and sorting parameters entry | The triggering surface of unauthenticated injection is just such an entry | Some public functions are temporarily unavailable |
| Second related | The reachable range of the background login page, limiting the source, and enabling captcha protection | Blocking credential brute force attacks and post-login operation chains | The operation team needs to enter through the specified entrance |
| Do it synchronously | Disable unnecessary extensions and themes, and delete the site instances that are no longer in use | Reduce the component surface that can be combined | Review the dependencies after the fact |
| Must do | Complete backup and verify recoverability | Can be rolled back if defeated, to avoid blocking of upgrades and recovery | Occupying window time |
The key to these two WordPress vulnerabilities is 'unauthenticated': It does not require login and relies on request parameters to enter the database query. For content management systems, it is worth prioritizing the entry points that are highly similar: sorting and filtering parameters on list pages, search interfaces, form submissions, import interfaces, and archive addresses with query parameters.
reCAPTCHA captcha and similar form protections cannot block injections, but they can block credential brute force and automated submission. They belong to concurrent actions on the same day and should not be considered as the main defense.
The correct sequence on patch day.
First verify the backup before making any changes to production. When upgrade failure and data anomalies caused by exploitation are mixed together, the only way to distinguish between them is with a usable backup.
Then judge the priority based on the scope of impact. When deploying across multiple sites, confirm the version of each instance individually, and do not assume that if the main site is upgraded, other sites will also be secure.
Do three things immediately after the upgrade: confirm that the version number has indeed changed; conduct a random check of the list, search, form, and background login; mark the suspicious requests before the upgrade in the access log for subsequent investigation.
The last processing of credentials. The upgrade only fixes the code path, and credentials that have been stolen or guessed will not become invalid due to the upgrade. The upgrade suggestions for AnQiCMS v3.6.6 (2026-10-08) include such actions: this version fixes two high-risk issues of list sorting parameter injection and site switching login credentials that can be forged, and after the upgrade, it is required to rotate the site signature key and change the administrator password. Not rotating the key after the code is fixed is equivalent to changing the lock of the door while leaving the key outside.
Long-term confinement of injection surface within system design
The passive ability of patch day comes from architectural selection. Three long-term practices can reduce the number of emergencies in the future:
Firstly, reduce the assembly query during runtime. Parameterized queries and whitelist field validation can block the vast majority of injection surfaces. AnQiCMS lists the defense against SQL injection and XSS, built-in JWT authentication, content sensitive word filtering, and anti-crawling interference codes as built-in capabilities in the security description of the README file, which belongs to the system level instead of plugin叠加.
Secondly, reduce the third-party aspect. Unauthenticated injection often appears in extensions and themes, and the more extensions you have, the longer the patch response chain you depend on.
The third level closes high-risk operations by default. Site-wide operations (backup, upgrade, multi-site management) are not exposed to the outside by default and can be explicitly enabled as needed, which can avoid the direct access to site-wide actions after a single entry is breached.
Common questions
Can we only close the background and not the public interface?No. The trigger points of the authentication-free vulnerabilities do not need the background in the first place, and closing the background has almost no harmful effect on such vulnerabilities.
Not enough time to upgrade, are the WAF rules sufficient?It can争取 time, but cannot replace the upgrade. After the public exploitation code spreads, rules bypassing appear very quickly, and WAF should be used as a buffer rather than a conclusion.
How can we determine if it has been exploited?Check the access logs before the upgrade for any abnormal long requests related to sorting/search parameters, database error pages, unknown account or permission changes in the background, or appearance of unfamiliar files in static files. If an anomaly is found, preserve the evidence first before cleaning up.