Hacker News Reader

Comment by cies | Hacker News Reader

PHP is know for being very leaky about being PHP (and since PHP + ecosystem have a bad history of CVEs, being leaky about being PHP is not cool).

Java/Kotlin/Go/Rust/Ruby/Python/JS/TS are a lot less leaky about what language the server-side is written in. Usually the webserver used advertises itself name and in a server string, but it is considered bad practice and thus often switched off.

Reading "php" extensions in paths is a clear giveaway, so are "htm" extensions for microsoft products. Tools usually guess the language/framework based on some of these giveaways and the better the tools the less this is evident.

I jut checked some web apps I worked on, and only the one I last touched 10+ years ago is detected with buildwith.com; it's a Rails site.

All the Java/Kotlin/Rust/Hasura+Elm apps I worked on since are now shown as "Nginx" (the rev proxy in front of it).

View on HN

I just checked the day gig site... builtwith claims it's using Webflow and Apollo GraphQL (Neither of which it is) and doesn't mention at all the language it's actually implemented in (Python), although that's not surprising since it's an in house framework.

How is being “leaky” in any way bad or even good?

Every programming language has holes, its just that with PHP the attack surface is much larger, so i guess people find more holes, etc..

Are you advocating “security by obscurity”?

While it kind of is security by obscurity, it's a very basic piece of server hardening to stop telling potential attackers what software you're using (within reason).

Back in the day (!), server software used to honestly respond with things like the software name and exact version number it was running.

Naturally, that meant scanning for vulnerabilities was a lot easier than it needed to be.

All security is by obscurity. Some is useful.

Not true. There are some real cryptographic realities that are based in "open" math principles.

There's also the way of most using runtimes/libraries that (constantly) have CVEs in them; and understanding why it is that these languages have CVEs in the first place (see my comment on "eval()").

A cryptographic key is an obscured secret.

Also, the overly muscled guys out the front of night clubs aren't there for "obscurity" type security. ;)

obviously being properly secure is better. but if you leave your unlocked, it's better to not also hang a sign above it saying "this door is unlocked".

obscurity is absolutely part of good security practice, as long as it's not all you're relying on.

> How is being “leaky” in any way bad or even good?

Information-gathering is a common early step in any attack against a system; knowing the language & libraries involved (especially their versions) allows you to search for any existing CVEs that apply.

> Are you advocating “security by obscurity”?

I don't think OP was implying that security by obscurity alone is sufficient, just that it's unwise to advertise information that's not relevant to end users, that could help would-be attackers.

If you languages has "eval()" or something similar, it is a lot easier to attack. Same for when it allow you "upload a file in some place where it gets executed".

These things are not so easy, say, with a C++/Rust/Go app. Or even in most JVM configurations. JS has similar issues, that Deno is trying to mitigate to some extend.

Reply on news.ycombinator.com