vLLM Vulnerable to Remote Code Execution via Mooncake Integration
Impacted Deployments
Note that vLLM instances that do NOT make use of the mooncake integration are NOT vulnerable.
Description
vLLM integration with mooncake is vaulnerable to remote code execution due to using pickle based serialization over unsecured ZeroMQ sockets. The vulnerable sockets were set to listen on all network interfaces, increasing the likelihood that an attacker is able to reach the vulnerable ZeroMQ sockets to carry out an attack.
This is a similar to GHSA - x3m8 - f7g5 - qhm7, the problem is in
https://github.com/vllm-project/vllm/blob/32b14baf8a1f7195ca09484de3008063569b43c5/vllm/distributed/kv_transfer/kv_pipe/mooncake_pipe.py#L179
Here recv_pyobj() Contains implicit pickle.loads(), which leads to potential RCE.
Data exposure via ZeroMQ on multi-node vLLM deployment
Impact
In a multi-node vLLM deployment, vLLM uses ZeroMQ for some multi-node communication purposes. The primary vLLM host opens an XPUB ZeroMQ socket and binds it to ALL interfaces. While the socket is always opened for a multi-node deployment, it is only used when doing tensor parallelism across multiple hosts.
Any client with network access to this host can connect to this XPUB socket unless its port is blocked by a firewall. Once connected, these arbitrary clients will receive all of the same data broadcasted to all of the secondary vLLM hosts. This data is internal vLLM state information that is not useful to an attacker.
By potentially connecting to this socket many times and not reading data published to them, an attacker can also cause a denial of service by slowing down or potentially blocking the publisher.
Detailed Analysis
The XPUB socket in question is created here:
https://github.com/vllm-project/vllm/blob/c21b99b91241409c2fdf9f3f8c542e8748b317be/vllm/distributed/device_communicators/shm_broadcast.py#L236-L237
Data is published over this socket via MessageQueue.enqueue() which is called by MessageQueue.broadcast_object():
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/device_communicators/shm_broadcast.py#L452-L453
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/device_communicators/shm_broadcast.py#L475-L478
The MessageQueue.broadcast_object() method is called by the GroupCoordinator.broadcast_object() method in parallel_state.py:
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/parallel_state.py#L364-L366
The broadcast over ZeroMQ is only done if the GroupCoordinator was created with use_message_queue_broadcaster set to True:
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/parallel_state.py#L216-L219
The only case where GroupCoordinator is created with use_message_queue_broadcaster is the coordinator for the tensor parallelism group:
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/parallel_state.py#L931-L936
To determine what data is broadcasted to the tensor parallism group, we must continue tracing. GroupCoordinator.broadcast_object() is called by GroupCoordinator.broadcoast_tensor_dict():
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/parallel_state.py#L489
which is called by broadcast_tensor_dict() in communication_op.py:
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/distributed/communication_op.py#L29-L34
If we look at _get_driver_input_and_broadcast() in the V0 worker_base.py, we'll see how this tensor dict is formed:
https://github.com/vllm-project/vllm/blob/790b79750b596043036b9fcbee885827fdd2ef3d/vllm/worker/worker_base.py#L332-L352
but the data actually sent over ZeroMQ is the metadata_list portion that is split from this tensor_dict. The tensor parts are sent via torch.distributed and only metadata about those tensors is sent via ZeroMQ.
https://github.com/vllm-project/vllm/blob/54a66e5fee4a1ea62f1e4c79a078b20668e408c6/vllm/distributed/parallel_state.py#L61-L83
Patches
https://github.com/vllm-project/vllm/pull/17197
Workarounds
Prior to the fix, your options include:
1. Do not expose the vLLM host to a network where any untrusted connections may reach the host.
2. Ensure that only the other vLLM hosts are able to connect to the TCP port used for the XPUB socket. Note that port used is random.
References
Relevant code first introduced in https://github.com/vllm-project/vllm/pull/6183
Transformers Regular Expression Denial of Service (ReDoS) vulnerability
A Regular Expression Denial of Service (ReDoS) vulnerability was identified in the huggingface/transformers library, specifically in the file tokenization_gpt_neox_japanese.py of the GPT-NeoX-Japanese model. The vulnerability occurs in the SubWordJapaneseTokenizer class, where regular expressions process specially crafted inputs. The issue stems from a regex exhibiting exponential complexity under certain conditions, leading to excessive backtracking. This can result in high CPU usage and potential application downtime, effectively creating a Denial of Service (DoS) scenario. The affected version is v4.48.1 (latest).
React Router allows pre-render data spoofing on React-Router framework mode
Summary
After some research, it turns out that it's possible to modify pre-rendered data by adding a header to the request. This allows to completely spoof its contents and modify all the values of the data object passed to the HTML. Latest versions are impacted.
Details
The vulnerable header is X-React-Router-Prerender-Data, a specific JSON object must be passed to it in order for the spoofing to be successful as we will see shortly. Here is the vulnerable code :
<img width="776" alt="Capture d’écran 2025-04-07 à 05 36 58" src="https://github.com/user-attachments/assets/c95b0b33-15ce-4d30-9f5e-b10525dd6ab4" />
To use the header, React-router must be used in Framework mode, and for the attack to be possible the target page must use a loader.
Steps to reproduce
Versions used for our PoC:
"@react-router/node": "^7.5.0",
"@react-router/serve": "^7.5.0",
"react": "^19.0.0"
"react-dom": "^19.0.0"
"react-router": "^7.5.0"
1. Install React-Router with its default configuration in Framework mode (https://reactrouter.com/start/framework/installation)
2. Add a simple page using a loader (example: routes/ssr)
3. Access your page (which uses the loader) by suffixing it with .data. In our case the page is called /ssr:
!image
We access it by adding the suffix .data and retrieve the data object, needed for the header:
!image
4. Send your request by adding the X-React-Router-Prerender-Data header with the previously retrieved object as its value. You can change any value of your data object (do not touch the other values, the latter being necessary for the object to be processed correctly and not throw an error):
!Capture d’écran 2025-04-07 à 05 56 10
As you can see, all values have been changed/overwritten by the values provided via the header.
Impact
The impact is significant, if a cache system is in place, it is possible to poison a response in which all of the data transmitted via a loader would be altered by an attacker allowing him to take control of the content of the page and modify it as he wishes via a cache-poisoning attack. This can lead to several types of attacks including potential stored XSS depending on the context in which the data is injected and/or how the data is used on the client-side.
Credits
Rachid Allam (zhero;)
Yasser Allam (inzo_)
React Router allows a DoS via cache poisoning by forcing SPA mode
Summary
After some research, it turns out that it is possible to force an application to switch to SPA mode by adding a header to the request. If the application uses SSR and is forced to switch to SPA, this causes an error that completely corrupts the page. If a cache system is in place, this allows the response containing the error to be cached, resulting in a cache poisoning that strongly impacts the availability of the application.
Details
The vulnerable header is X-React-Router-SPA-Mode; adding it to a request sent to a page/endpoint using a loader throws an error. Here is the vulnerable code :
<img width="672" alt="Capture d’écran 2025-04-07 à 08 28 20" src="https://github.com/user-attachments/assets/0a0e9c41-70fd-4dba-9061-892dd6797291" />
To use the header, React-router must be used in Framework mode, and for the attack to be possible the target page must use a loader.
Steps to reproduce
Versions used for our PoC:
"@react-router/node": "^7.5.0",
"@react-router/serve": "^7.5.0",
"react": "^19.0.0"
"react-dom": "^19.0.0"
"react-router": "^7.5.0"
1. Install React-Router with its default configuration in Framework mode (https://reactrouter.com/start/framework/installation)
2. Add a simple page using a loader (example: routes/ssr)
!image
3. Send a request to the endpoint using the loader (/ssr in our case) adding the following header:
``
X-React-Router-SPA-Mode: yes
``
Notice the difference between a request with and without the header;
Normal request
!Capture d’écran 2025-04-07 à 08 36 27
With the header
!Capture d’écran 2025-04-07 à 08 37 01
!image
Impact
If a system cache is in place, it is possible to poison the response by completely altering its content (by an error message), strongly impacting its availability, making the latter impractical via a cache-poisoning attack.
Credits
Rachid Allam (zhero;)
Yasser Allam (inzo_)
vLLM allows Remote Code Execution by Pickle Deserialization via AsyncEngineRPCServer() RPC server entrypoints
vllm-project vllm version 0.6.0 contains a vulnerability in the AsyncEngineRPCServer() RPC server entrypoints. The core functionality run_server_loop() calls the function _make_handler_coro(), which directly uses cloudpickle.loads() on received messages without any sanitization. This can result in remote code execution by deserializing malicious pickle data.
OWASP A03OWASP A08LLM01 · Prompt InjectionLLM05 · Supply Chain
Get guardrail →vLLM deserialization vulnerability in vllm.distributed.GroupCoordinator.recv_object
vllm-project vllm version 0.6.0 contains a vulnerability in the distributed training API. The function vllm.distributed.GroupCoordinator.recv_object() deserializes received object bytes using pickle.loads() without sanitization, leading to a remote code execution vulnerability.
Maintainer perspective
Note that vLLM does NOT use the code as described in the report on huntr. The problem only exists if you use these internal APIs in a way that exposes them to a network as described. The vllm team was not involved in the analysis of this report and the decision to assign it a CVE.
Gradio DOS in multipart boundry while uploading the file
A vulnerability in the file upload process of gradio-app/gradio version @gradio/video@0.10.2 allows for a Denial of Service (DoS) attack. An attacker can append a large number of characters to the end of a multipart boundary, causing the system to continuously process each character and issue warnings. This can render Gradio inaccessible for extended periods, disrupting services and causing significant downtime.
Gradio Vulnerable to Open Redirect
An open redirect vulnerability exists in the latest version of gradio-app/gradio. The vulnerability allows an attacker to redirect users to a malicious website by URL encoding. This can be exploited by sending a crafted request to the application, which results in a 302 redirect to an attacker-controlled site.
LlamaIndex vulnerable to Creation of Temporary File in Directory with Insecure Permissions
A vulnerability in the default_jsonalyzer function of the JSONalyzeQueryEngine in the run-llama/llama_index repository allows for SQL injection via prompt injection. This can lead to arbitrary file creation and Denial-of-Service (DoS) attacks. The vulnerability affects the latest version and is fixed in version 0.12.3.
LlamaIndex Uncontrolled Resource Consumption vulnerability
A vulnerability in the KnowledgeBaseWebReader class of the run-llama/llama_index repository, version latest, allows an attacker to cause a Denial of Service (DoS) by controlling a URL variable to contain the root URL. This leads to infinite recursive calls to the get_article_urls method, exhausting system resources and potentially crashing the application.
Transformers Regular Expression Denial of Service (ReDoS) vulnerability
A Regular Expression Denial of Service (ReDoS) vulnerability was identified in the huggingface/transformers library, specifically in the file tokenization_nougat_fast.py. The vulnerability occurs in the post_process_single() function, where a regular expression processes specially crafted input. The issue stems from the regex exhibiting exponential time complexity under certain conditions, leading to excessive backtracking. This can result in significantly high CPU usage and potential application downtime, effectively creating a Denial of Service (DoS) scenario. The affected version is v4.46.3.
LlamaIndex Improper Handling of Exceptional Conditions vulnerability
A vulnerability in the LangChainLLM class of the run-llama/llama_index repository, version v0.12.5, allows for a Denial of Service (DoS) attack. The stream_complete method executes the llm using a thread and retrieves the result via the get_response_gen method of the StreamingGeneratorCallbackHandler class. If the thread terminates abnormally before the _llm.predict is executed, there is no exception handling for this case, leading to an infinite loop in the get_response_gen function. This can be triggered by providing an input of an incorrect type, causing the thread to terminate and the process to continue running indefinitely.
Gradio Path Traversal vulnerability
A vulnerability in the gradio-app/gradio repository, version git 67e4044, allows for path traversal on Windows OS. The implementation of the blocked_path functionality, which is intended to disallow users from reading certain files, is flawed. Specifically, while the application correctly blocks access to paths like 'C:/tmp/secret.txt', it fails to block access when using NTFS Alternate Data Streams (ADS) syntax, such as 'C:/tmp/secret.txt::$DATA'. This flaw can lead to unauthorized reading of blocked file paths.
vLLM Deserialization of Untrusted Data vulnerability
vllm-project vllm version v0.6.2 contains a vulnerability in the MessageQueue.dequeue() API function. The function uses pickle.loads to parse received sockets directly, leading to a remote code execution vulnerability. An attacker can exploit this by sending a malicious payload to the MessageQueue, causing the victim's machine to execute arbitrary code.
langchain-core allows unauthorized users to read arbitrary files from the host file system
A vulnerability in langchain-core versions >=0.1.17,<0.1.53, >=0.2.0,<0.2.43, and >=0.3.0,<0.3.15 allows unauthorized users to read arbitrary files from the host file system. The issue arises from the ability to create langchain_core.prompts.ImagePromptTemplate's (and by extension langchain_core.prompts.ChatPromptTemplate's) with input variables that can read any user-specified path from the server file system. If the outputs of these prompt templates are exposed to the user, either directly or through downstream model outputs, it can lead to the exposure of sensitive information.
Gradio Vulnerable to Arbitrary File Deletion
A path traversal vulnerability exists in the Gradio Audio component of gradio-app/gradio, as of version git 98cbcae. This vulnerability allows an attacker to control the format of the audio file, leading to arbitrary file content deletion. By manipulating the output format, an attacker can reset any file to an empty file, causing a denial of service (DOS) on the server.
Gradio Vulnerable to Denial of Service (DoS) via Crafted HTTP Request
A Regular Expression Denial of Service (ReDoS) vulnerability exists in the gradio-app/gradio repository, affecting the gr.Datetime component. The affected version is git commit 98cbcae. The vulnerability arises from the use of a regular expression ^(?:\snow\s(?:-\s(\d+)\s([dmhs]))?)?\s*$ to process user input. In Python's default regex engine, this regular expression can take polynomial time to match certain crafted inputs. An attacker can exploit this by sending a crafted HTTP request, causing the gradio process to consume 100% CPU and potentially leading to a Denial of Service (DoS) condition on the server.
Gradio Vulnerable to Denial of Service (DoS) via Crafted Zip Bomb
A vulnerability in the dataframe component of gradio-app/gradio (version git 98cbcae) allows for a zip bomb attack. The component uses pd.read_csv to process input values, which can accept compressed files. An attacker can exploit this by uploading a maliciously crafted zip bomb, leading to a server crash and causing a denial of service.
vLLM Allows Remote Code Execution via Mooncake Integration
Summary
When vLLM is configured to use Mooncake, unsafe deserialization exposed directly over ZMQ/TCP will allow attackers to execute remote code on distributed hosts.
Details
1. Pickle deserialization vulnerabilities are well documented.
2. The mooncake pipe is exposed over the network (by design to enable disaggregated prefilling across distributed environments) using ZMQ over TCP, greatly increasing exploitability. ~~Further, the mooncake integration opens these sockets listening on all interfaces on the host, meaning it can not be configured to only use a private, trusted network.~~
Only sender_socket and receiver_ack are allowed to be accessed publicly, while the data actually decompressed by pickle.loads() comes from recv_bytes. Its interface is defined as self.receiver_socket.connect(f\"tcp://{d_host}:{d_rank_offset + 1}\"), where d_host is decode_host, a locally defined address 192.168.0.139,from mooncake.json (https://github.com/kvcache-ai/Mooncake/blob/main/doc/en/vllm-integration-v0.2.md?plain=1#L36).
3. The root problem is recv_tensor() calls _recv_impl which passes the raw network bytes to pickle.loads(). Additionally, it does not appear that there are any controls (network, authentication, etc) to prevent arbitrary users from sending this payload to the affected service.
Impact
This is a remote code execution vulnerability impacting any deployments using Mooncake to distribute KV across distributed hosts.
Remediation
This issue is resolved by https://github.com/vllm-project/vllm/pull/14228
