« first day (3416 days earlier)      last day (1762 days later) » 

11:06
0
Q: Can the Host Header be used to hide the existence of a service?

MechMK1Imagine a web server running on 93.184.216.34, usually reachable via the public DNS entry example.com. Web servers usually allow the distinction of multiple "virtual" servers, based on the Host header received via the HTTP request. Now imagine the same web server would offer a different service ...

 
2 hours later…
12:39
@MechMK1 interesting case... and I believe it works better than most people think
Yeah. I mean, as long as it doesn't go via public dns, it should be pretty obscure
even if it does, if I ask my dns server about your domain, it will return NXDOMAIN. leaks the hostname but not the ip
kinda like leaking a password but not the username
Why? If you ask for example.com, you get a regular A record
the "secret" host never hits DNS at all. It'd just be done via the hosts file
if I ask for cx6wdffpuik997eljf6d878i6f3np4207ne30vyjsvhpra69 it returns NXDOMAIN
exactly, because it is not a valid domain
12:46
but on the dns log, there will an interesting entry about a domain cx6wdffpuik997eljf6d878i6f3np4207ne30vyjsvhpra69
True that
an admin locked at home with lots of free time may wonder why he is getting that strange name over and over... ask around, see your question, my answer...
what about scanning the whole internet?
What do you mean?
Oh, you mean an admin trying it?
Well, he might then reach the target, but with invalid certificates
:D
if you do that with my server, it forwards you to my question
only works with http though
client certificates would protect it, but would do the same for a common domain too
if you use client certificates, the obscure domain not add much security
It's the security by design and some obscurity sprinkled ontop
Like disabling server headers
Doesn't protect you, but doesn't hurt you either
12:59
it's like having a secret door with a very secure lock, and a visible door with a very secure lock
Yeah, but the visible door springs to view immediately
if the door is visible or not is not what makes it secure, but the strong lock
The secrecy of the door may be a feature in itself
it's kinda like my personal server. ssh is not on port 22. it does not make it more secure, only decreases the automated attacks. if someone bruteforces my server, it will someday find the correct port
the security is from not being able to login with passwords, and I strapped googleauth pam module on it too, so the attacker would have to get my key AND my 2FA token
changing the port does little on the security
it even hurts me some times when I forget to change the port while connecting to it and ssh hangs waiting for the connection
It's even worse if you have a honeypot on SSH 22
Which does nothing except receive auth requests and give veeeery slow responses
13:05
in one server, I do (sshpot). and I have portsentry. and fail2ban.
I'm just a pleb. SSH on port 22 and pubkey auth
I use sslh (or sshl, I never remember) and it stays in port 443... so if I send an HTTPS request, it sends it to Apache. if it's an openvpn, it redirects to openvpn. if I ssh into it, it sends to sshd...
it's a nice addition, as most proxies allow CONNECT on port 443... little they know...
what the fuck is this abomination
Back in my days we used to have dedicated ports per protocol and WE LIKED IT!
Kids these days, no respect for traditions
13:08
About the web site question, ultimately you can put an auth token pretty much anywhere and it has a similar effect
My labs use an auth token in the host name, with wildcard DNS and SSL
@paj28 Yeah, I saw that being done with a cookie once
why dedicate ports if you can use only one port and have everything?
I didn't want to use cookies because they fiddle with cookies as part of the labs, so it could be confusing
If you had secret_feature=hfojshqjewtwjdvnsjbkjybxlgajgjaljhg, it would enable some secret features
Putting the token in the host name is the least aggro
13:09
I want that too
Just make a website that is full of easter eggs
My friend got a trip to New York over something similar
Then in the end start a massive ARG, all while denying that an ARG is happening
To the future reader: Welcome. There is no ARG. Go home.
I'll have to look at it
 
8 hours later…
Luc
Luc
21:25
I think you'd have to accept a bad edit then flag the edited question. It's probaly rare, but slight SE design flaw. Should be a link to flag the edits.
I chatted to a db security guy earlier. We got onto the issue of smart db vs smart app, dumb db.
Of course, as a db sec guy he has a vested interest in the smart db model
He made an interesting point that even if you have an app front-end for end users
You'll have data scientists and people with partial administrative who need direct access
21:47
@Luc Mod flag one of the questions and link to the suggested edit
Here's what they say on Stack Overflow about that type of situation: meta.stackoverflow.com/a/360812
I'm guessing they have the same policy here
It looks like Schroeder already rejected one of the edits anyway

« first day (3416 days earlier)      last day (1762 days later) »