When using the xhtml1-transitional.dtd
doctype, collecting a credit card number with the following HTML
<input type="text" id="cardNumber" name="cardNumber" autocomplete='off'/>
will flag a warning on the W3C validator:
there is no attribute "autocomplete".
Is there a W3C / standards way to disable browser auto-complete on sensitive fields in a form?
autocomplete="off"
this should fix the issue for all modern browsers.In current versions of Gecko browsers, the autocomplete attribute works perfectly. For earlier versions, going back to Netscape 6.2, it worked with the exception for forms with "Address" and "Name"
Update
In some cases, the browser will keep suggesting autocompletion values even if the autocomplete attribute is set to off. This unexpected behavior can be quite puzzling for developers. The trick to really forcing the no-autocompletion is to assign a random string to the attribute, for example:
Since this random value is not a
valid one
, the browser will give up.Documetation
Another way - which will also help with security is to call the input box something different every time you display it: just like a captha. That way, the session can read the one-time only input and Auto-Complete has nothing to go on.
Just a point regarding rmeador's question of whether you should be interfering with the browser experience: We develop Contact Management & CRM systems, and when you are typing other people's data into a form you don't want it constantly suggesting your own details.
This works for our needs, but then we have the luxury of telling users to get a decent browser:)
Using a random 'name' attribute works for me.
I reset the name attribute when sending the form so you can still access it by name when the form is sent. (using the id attribute to store the name)
I think there's a simpler way. Create a hidden input with a random name (via javascript) and set the username to that. Repeat with the password. This way your backend script knows exactly what the appropriate field name is, while keeping autocomplete in the dark.
I'm probably wrong, but it's just an idea.
I suggest catching all 4 types of input:
Reference: