Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
Tools/GitHubGitHub/kinokopio/cve-2026-23111
Privilege EscalationVulnerability AnalysisExploitationPost-ExploitationPayload DevelopmentContainer EscapeBinary Exploitation
GitHubkinokopio/cve-2026-23111

CVE-2026-23111

nf_tables catchall UAF exploit for Ubuntu 6.8.0-101-generic: unprivileged local privilege escalation to host root and container escape via forged nsproxy and fs.

View Repository
611 month agoNot yet reviewed
Share

CVE-2026-23111 — Privilege Escalation & Container Escape PoC

nf_tables catchall UAF exploit for Ubuntu 6.8.0-101-generic:

  • Local privilege escalation: unprivileged user → real host root
  • Container escape: container uid 1000 → host root, with the host filesystem and namespaces

中文文档 / Chinese

For authorized test environments only. A successful run leaves the host kernel in an unrecoverable corrupted state — reboot when done.

Vulnerability

An inverted genmask check in nft_map_catchall_activate(): during a transaction abort the handler skips the inactive catchall elements that need reactivation and processes the active ones, so chain->use is decremented without restoration. DELCHAIN can then free a chain that is still referenced by a verdict — a use-after-free.

Affected branches: 5.10.188+, 5.15.121+, 6.1.36+, 6.7–6.12.69, 6.13–6.18.9.

Capabilities

Privilege escalation (lpe mode): the ROP chain overwrites the target task's real_cred and cred with init_cred, giving real uid 0, full capabilities, and lifting the LSM restriction. Works on bare metal or directly on the host.

Container escape (default mode): on top of that it overwrites nsproxy (a forged one, switching mnt/uts/ipc/net/time/cgroup to the host's) and fs (init_fs, so path resolution starts at the host root without setns() — the docker-default seccomp does not block it). After success the host /tmp holds a setuid rootsh and you get a host shell.

Both modes ship in one binary and share the same exploit chain and ROP; only the number of overwritten fields differs.

Target environment

ItemValue
OSUbuntu 22.04.5 LTS
Kernel6.8.0-101-generic (offsets are hardcoded for this build)
ContainerDocker, --user 1000:1000, --security-opt seccomp=unconfined
Hostkernel.unprivileged_userns_clone=1, nf_tables available

Build

make

Usage

On the test host the project is deployed at /home/ubuntu/poc-cve-2026-23111 (with exploit_escape2 already built).

Local privilege escalation

# run as an unprivileged user
cd /home/ubuntu/poc-cve-2026-23111
./exploit_escape2 check                                 # vulnerability check
./exploit_escape2 lpe                                   # interactive root shell
./exploit_escape2 lpe 'id; head -2 /etc/shadow'         # one-shot command

LPE: uid 1000 → uid 0

Container escape

# mount the project directory read-only into the container
docker run -d --name victim --user 1000:1000 \
  --security-opt seccomp=unconfined \
  -v /home/ubuntu/poc-cve-2026-23111:/poc:ro \
  mirror.ccs.tencentyun.com/library/ubuntu:22.04 sleep infinity

docker exec victim /poc/exploit_escape2 check           # vulnerability check
docker exec -it victim /poc/exploit_escape2             # interactive host root shell
docker exec victim /poc/exploit_escape2 'id; head -2 /etc/shadow'   # one-shot command

Escape: container uid 1000 → host root

Implementation notes

  • A parked escapee child process is the overwrite target; after the ROP the hijacked thread only calls pause() — no business logic runs in the hijacked context.
  • Both modes share one ROP: lpe writes real_cred/cred; escape adds nsproxy/fs.
  • The escapee never exits (it runs on init objects written without matching refcounts) and ignores every catchable fatal signal.
  • The escaped shell is forked by the escapee and stays in the container pid ns, so fork/wait/reap behave normally.
  • Post-trigger failures park instead of exiting, so no cleanup ever runs on the corrupted netns.

Limitations

  1. The kernel must be exactly 6.8.0-101-generic; other flavors or point releases require re-extracting the offsets.
  2. Success rate is roughly 50% (probabilistic slab reclaim); a miss panics the host when panic_on_oops=1.
  3. Even on success the host stays corrupted and must be rebooted; containers using the default seccomp profile cannot be exploited.

Operational rules

  1. Never kill esc_tgtX / nftpwnAB / parked threads, and never docker rm -f a container that ran the exploit.
  2. The only safe cleanup is reboot.
  3. If the session hangs, close the terminal and open a new SSH session; do not kill %1.

References

  • Upstream PoC and offset-extraction tooling: https://github.com/Knz-source/CVE-2026-23111-POC-noddlenpottato
  • Container-escape architecture reference (parked escapee / e2e): https://github.com/ctrsploit/ctrsploit/tree/main/vul/cve-2026-23111
  • Vulnerable code: net/netfilter/nf_tables_api.c — nft_map_catchall_activate()
Download Tool

Most Popular

View all →

Discover the most used tools by our community.

No popularity data yet.

Explore all tools

Browse our collection of tools

View all tools →