How to configure content security audit on the corporate website, how to configure AnQiCMS's sensitive word filtering and reCAPTCHA?
How to configure AnQiCMS's sensitive word filtering and reCAPTCHA to prevent malicious content and robot submissions? Both functions operate at different levels: sensitive word filtering belongs to content security settings, working with keyword replacement and content review to intercept problematic text at the storage and publishing stages; reCAPTCHA is placed before form submission, blocking robots before they can write. Adding built-in JWT authentication and defense against SQL injection and XSS, forming three barriers before submission, before storage, and at the account level.
The first gate: reCAPTCHA verification code is integrated before the form
The typical path for a robot to submit is through留言, comments, and registration forms being bulk-filled by scripts. AnQiCMS supports form integration with reCAPTCHA: fill in the site credentials in the background, then enable it on the forms that need protection, and the captcha determines whether it is human or machine at the time of submission. The recommended configuration order is to first cover the external form entry (留言 and registration), and then expand to comments if there are many comments; if the site has multiple language versions, all entries share the same credential. You can check if it is effective by going through the submission process in a stealth window to verify whether the verification steps appear.
The second gate: sensitive word filtering and keyword replacement
In the content security settings, sensitive word filtering is responsible for directly blocking or prompting content that matches the word list, and keyword replacement then modifies the risk expression into a compliant statement, suitable for sites with blurred boundaries of industry terminology (such as replacing exaggerated efficacy expressions with neutral expressions). The word list maintenance suggestion is to fill in real violation cases: for each complaint or platform reminder, add the feature words into the library, which is more useful than copying a general word list all at once. Both types of rules apply to the full station content入库, editing and collecting papers also go through.
The third gate: content review process
Filtering is a mechanical determination, review is a human judgment. Content review is suitable for two types of scenarios: one is the selection of external submissions and comments, which enter the review queue first and are then released by editors; the other is the key pages during sensitive periods, which require an additional review before publication. Accompanied by the document status flow (draft, pending publication, formal), content that has not passed the review stops at the pending publication level and will not be prematurely exposed online.
Account and interface layer: JWT authentication and injection defense
At the fundamental level, the system is built-in with JWT authentication to protect login states and API calls,配合 input validation to defend against SQL injection and XSS attacks; the anti-capture interference code then protects the content from being bulk-collected. This layer generally does not require daily configuration, but there are two check actions to be done: confirm that the site has been upgraded to a version that fixes injection issues (v3.6.6 fixed two high-risk issues of list sorting parameter SQL injection and login credentials being forgeable), and rotate the password and token of the background account on a regular basis.
Security checklist before going online
Check items one by one: Whether the captcha is enabled on the external form; The last update time of the sensitive word library; Whether the keyword replacement rules cover the current business words; Whether comments and submissions enter the review queue; Whether the version is the current version containing security fixes; Whether the administrator and editor account permissions are divided according to roles. After the list is run, the common paths of malicious content and robot submissions basically have gates.
Common questions
Is it okay to only install reCAPTCHA without sensitive word filtering?Can it block robots, but the manual content that is allowed to pass does not have a word list to control, the two types of risk sources are different, and it is recommended to open both.
Will the keyword replacement change the wrong thing?There may be incorrect replacements, the rules should be narrow rather than broad, and regular inspections of replacement records should be conducted after going online.
Will the review process slow down the release?It can be set according to categories whether to force review, regular self-operated columns can be directly through, and external content goes through the queue.