The recent post about what people are using for webmail got me thinking about a perhaps irrational policy I have with my own self-hosted software: I don’t install anything written in PHP, because I have this vague notion that PHP software is often insecure. I think I probably got this idea because years ago I saw all the vulnerabilities in PHP webmail clients and PHP software like Wordpress and decided that it was the language’s fault—or at least a contributing factor.
Maybe this isn’t fair. Maybe PHP is just more accessible to new devs and so they’re more likely to gravitate to it and make security mistakes. Maybe my perception isn’t even accurate, and webmail / blog software written in other languages is just as bad—but PHP gets all the the negative attention because it’s so prevalent for web apps. Maybe my policy was a good idea, years ago, but now it’s just out of date.
To be clear, I’m not trying to stoke the flames of a language holy war here or anything. I’m honestly asking: Is it maybe time to revisit my anti-PHP policy? I’m looking longingly at some federated software like Pixelfed and wondering if maybe I’m just being a little too close-minded.
So I’m interested in your own experiences and polices here. Where do you draw the security line for what you will or won’t host, and what made you make that choice?
I don’t think programming language is a good metric for security. I assume everything I host has issues, and then try to mitigate from there.
IMHO, a better approach is to vet the project beforehand, looking at whether it is still actively maintained. I usually use things like commits, issues, etc to try and gauge whether a piece of software is actively maintained so that when an issue arises, it can be fixed.
You can mitigate much of the risk by using some basic best practices, like isolating all apps from each other (using docker, for example), using a reverse proxy, tools like fail2ban or a web application firewall, using proper database permissions for each app, etc
What I also do is add another layer by making certain applications accessible only over vpn. That won’t work for some tools, obviously, but also reduces the risk for tools you are only using yourself.
That all seems prudent and reasonable. I guess some of my own anxiety is about how exactly I’ll evaluate projects like you’re talking about. I can (and do) certainly look at whether a project is actively developed before selecting it. Not just for security reasons… I don’t want to bet on a horse that won’t get updated with fixes and features. But for security in particular, I guess I was hoping for ways to evaluate that for a project… without exhaustively poring over its source. Maybe, to your point, the other mitigations you listed should be sufficient, and I should worry more about that side of things than picking the perfect project.