> Not really. Having the ability for a web process to write its to files which it then executes actually opens up whole new vulnerabilities. For example: one popular PHP exploit is writing a PHP file, then executing it on the remote system. Can't write a file? Can't execute the exploit.
You're definitely correct, but there can be two different contexts to execute PHP; with differing permissions. If I was to rebuild how WordPress does it's autoupdating, I'd get rid of the god-awful HTTP "cron" crap, and have a proper PHP CLI that can do the updating, that runs on the server and is never exposed outside of it, and then WordPress itself, with the core outside the webroot and the uploads folder with zero executable permissions.
Still not perfect, but it'd be a hell of a lot nicer than how it currently runs.
Re:In Defence of WordPress
by plorkyeran on Friday May 15, 2015 @03:07AM
(#9549291)
Replacing the self-updating with an external updater does eliminate the problems with self-updating, but it's not a viable option for wordpress's primary deployment target of terrible shared hosting.
Not really. Having the ability for a web process to write its to files which it then executes actually opens up whole new vulnerabilities. For example: one popular PHP exploit is writing a PHP file, then executing it on the remote system. Can't write a file? Can't execute the exploit.
Reading and executing PHP out of the DB, and also writing user input to that same database? Same problem. Triggering system-altering actions based on an unsigned token? That's a vulnerability.
Letting an application be able to do more on a system than is absolutely necessary to operate is a great way to give control and information over to others.