Feature preview

Feature Introduction

Sixty percent of the submissions on unattended forms come from robots, should we add a hidden field first or put up a captcha first?

The form without protection has more than 60% of submissions coming from robots, the order to prevent spam is to add hidden fields first and then consider the captcha: hidden fields do not disturb real people, can block a large number of basic robots filling out templates, but cannot block manual replacement services; the captcha improves the interception limit but also increases the probability of real people being mistakenly rejected. Decide whether to add a second layer based on the number of complaints and the ratio of escapees.

Form anti-spam Captcha

Feature Introduction

Forms that are not protected have more than 60% of submissions coming from robots, should we add a hidden field first or first add a captcha: add a hidden field first. The day's statistics show that more than 60% of contact forms without human supervision are automatically submitted, and the hidden field can block about 40% of basic robots that fill out templates in bulk; it does not change the real person's form filling experience and does not require external services, and belongs to the first layer of zero interference. The captcha is the second layer, which is added when the leakage rate of the hidden field is still high.

Align the two day's statistics first

The principle of hiding fields is to add an input item that is invisible to humans and filled by scripts. Once the field has a value, it is determined to be a robot. Its blocking limit depends on how

The difference at this level of captcha is that it introduces an external judgment. The contrasting statements verified on the day were: hidden fields do not need to display any challenges to humans, while the captcha with ratings will directly reject the submission when judged as a robot, and humans may be mistakenly rejected.

The order of stacking layers

Hierarchy Means Degree of disturbance What can it block When to go up
The first layer Hidden fields and time check None Basic robot filled in bulk according to the template The form was filled out on the day of launch
The second layer Frequency Limitation and Source Verification None or very small Single-point high-frequency submission and script retry When the amount of spam continues to be the majority
The third layer Captcha Yes, may be incorrectly rejected Medium level of automation Complaints caused by the first two layers escaping
The fourth layer Manual review and keyword filtering Indifferent to the submitter Fake plays and low-quality content The first three layers are not enough

If the order is wrong, it will be a waste of cost: put the captcha directly, the loss of real people filling in the form is immediately visible, while the fake traffic still comes in; do the first and second layers first, then decide the third layer according to the data, the form quantity of most sites does not need to reach the fourth layer.

Form and comments are two entrances

The anti-spam function for forms and comments in AnQiCMS is separated into two switches, and do not think that setting one place will take effect globally. On the comment side, in addition to the interception mechanism, it can also go through sensitive word filtering and keyword replacement to handle the incoming spam content at the display layer; on the form side, more attention is paid to the submission action itself, relying on captcha and hidden fields to control.

Keep a record that can be checked for the effect: Record the total number of submissions per day, the number of those identified as robots, and the number of false positives confirmed manually. It is easy to keep a certain type of real person (such as visitors using screen reading tools) out for a long time if only the number of interferences is counted and not the false positives.

Common questions

What: Will hidden fields be skipped by advanced scripts? Answer: Yes. Its positioning is as the first layer with low cost, and the traffic skipping it needs to rely on frequency restrictions and manual review.

What: If a captcha is used, is it necessary to hide fields? Answer: It is not recommended. The two layers have different purposes. Hiding fields can reduce the traffic that reaches the captcha, thereby reducing the number of distractions caused by real people.

What: If the amount of spam is not large, do you need to add a captcha? Answer: Observe first. The loss of real person false rejection brought by captcha on low traffic sites is usually more expensive than the spam itself.

Related functions

Preview

The static site generator only produces pages, where should the edited content be changed?

The static site generator itself does not come with an editing entry, content modifications are made in the text files of the code repository, after modification, it must be regenerated and pushed to count as going live. Changes requiring engineering intervention are usually of three types: field structure adjustment, template modification, and batch replacement. A content management system with a backend incorporates these three steps into the interface, custom fields determine which level of editing can be modified, visual template editing and batch import reduce dependency on code, and a draft preview can be used before publication.

Preview

The custom content model has added a new field, will the text of old articles not be readable

The custom content model adds a new field, which only affects the field level, the text of the old articles will not be unreadable: the new field is empty in the existing documents, the template reads an empty value instead of reporting an error. The real problem is that the template treats empty values as required items, and when importing in bulk, the column names are not aligned with the new fields. Backup before changing, migrate in batches by category, and you can still go back if there is an error.

Preview

Found the word violation after the manuscript was sent out, where did the content review intercept?

After the manuscript was sent out, it was found that the wording was违规, the problem lies in placing all the compliance checks after publication. The sensitive word filter takes effect when written, content review takes effect before publication, keyword replacement and full-site replacement are used to batch change existing expressions. The three defenses are each set at different steps, in conjunction with the status transition of draft, to be published, recycling station, to ensure that illegal words do not go online, and if they do, they can also be taken back.