vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server
### Summary
A resource-exhaustion (denial-of-service) vulnerability exists in multiple endpoints of the OpenAI-Compatible Server due to the ability to specify Jinja templates via the `chat_template` and `chat_template_kwargs` parameters. If an attacker can supply these parameters to the API, they can cause a service outage by exhausting CPU and/or memory resources.
### Details
When using an LLM as a chat model, the conversation history must be rendered into a text input for the model. In `hf/transformer`, this rendering is performed using a Jinja template. The OpenAI-Compatible Server launched by vllm serve exposes a `chat_template` parame
xml2rfc is vulnerable to arbitrary file reads through prepped files
### Impact
When generating PDF files, this vulnerability allows an attacker to read arbitrary files from the filesystem by injecting malicious link element into the prepped RFCXML.
### Workarounds
Test untrusted input with `link` elements with `rel="attachment"` before processing.
### References
This is related to [GHSA-cfmv-h8fx-85m7](https://github.com/ietf-tools/xml2rfc/security/advisories/GHSA-cfmv-h8fx-85m7).
### Impact
When generating PDF files, this vulnerability allows an attacker to read arbitrary files from the filesystem by injecting malicious link element into the XML.
### Workarounds
Test untrusted input with `link` elements with `rel="attachment"` before processing.
### Credits
This vulnerability was reported by Mohamed Ouad from [Doyensec](https://doyensec.com/).
vLLM has remote code execution vulnerability in the tool call parser for Qwen3-Coder
### Summary
An unsafe deserialization vulnerability allows any authenticated user to execute arbitrary code on the server if they are able to get the model to pass the code as an argument to a tool call.
### Details
vLLM's [Qwen3 Coder tool parser](https://github.com/vllm-project/vllm/blob/main/vllm/entrypoints/openai/tool_parsers/qwen3coder_tool_parser.py) contains a code execution path that uses Python's `eval()` function to parse tool call parameters. This occurs during the parameter conversion process when the parser attempts to handle unknown data types.
This code path is reached when:
1. Tool calling is enabled (`--enable-auto-tool-c
**I. Detailed Description:**
This includes scenarios, screenshots, vulnerability reproduction methods. For account-related vulnerabilities, please provide test accounts. If the reproduction process is complex, you may record a video, upload it to Taopan, and attach the link.
1. Install ms-swift
```
pip install ms-swift -U
```
2. Start web-ui
```
swift web-ui --lang en
```
3. After startup, access through browser at [http://localhost:7860/](http://localhost:7860/) to see the launched fine-tuning framework program
4. Fill in necessary parameters
In the LLM Training interface, fill in required parameters including Mode
Part of the "Hades" wave of the Shai-Hulud supply-chain campaign. On 2026-06-08,
malicious phantom releases of instructor-mcp were published to PyPI using stolen
credentials. The package executes a bundled JavaScript payload (via the Bun
runtime) on import that harvests and exfiltrates credentials and attempts
self-propagation. This entry is a summary; behavior may not be fully
characterized here. See the linked references for detailed analysis and
indicators of compromise.
Part of the "Hades" wave of the Shai-Hulud supply-chain campaign. On 2026-06-08,
malicious phantom releases of dreamgen were published to PyPI using stolen
credentials. The package executes a bundled JavaScript payload (via the Bun
runtime) on import that harvests and exfiltrates credentials and attempts
self-propagation. This entry is a summary; behavior may not be fully
characterized here. See the linked references for detailed analysis and
indicators of compromise.
Open WebUI allows Remote Code Execution via Arbitrary File Upload to /audio/api/v1/transcriptions
OpenWebUI version 0.3.0 contains a vulnerability in the audio API endpoint `/audio/api/v1/transcriptions` that allows for arbitrary file upload. The application performs insufficient validation on the `file.content_type` and allows user-controlled filenames, leading to a path traversal vulnerability. This can be exploited by an authenticated user to overwrite critical files within the Docker container, potentially leading to remote code execution as the root user.
Open WebUI stored cross-site scripting (XSS) vulnerability
A stored cross-site scripting (XSS) vulnerability exists in open-webui/open-webui version 0.3.8. The vulnerability is present in the `/api/v1/models/add` endpoint, where the model description field is improperly sanitized before being rendered in chat. This allows an attacker to inject malicious scripts that can be executed by any user, including administrators, potentially leading to arbitrary code execution.
Open WebUI Has Improper Access Control Leading to Arbitrary Prompt Read
In version v0.3.8 of open-webui/open-webui, improper access control vulnerabilities allow an attacker to view any prompts. The application does not verify whether the attacker is an administrator, allowing the attacker to directly call the /api/v1/prompts/ interface to retrieve all prompt information created by the admin, which includes the ID values. Subsequently, the attacker can exploit the /api/v1/prompts/command/{command_id} interface to obtain arbitrary prompt information.
Open WebUI Cross-Site Request Forgery (CSRF) Vulnerability
A vulnerability in open-webui/open-webui versions <= 0.3.8 allows remote code execution by non-admin users via Cross-Site Request Forgery (CSRF). The application uses cookies with the SameSite attribute set to lax for authentication and lacks CSRF tokens. This allows an attacker to craft a malicious HTML that, when accessed by a victim, can modify the Python code of an existing pipeline and execute arbitrary code with the victim's privileges.
An improper access control vulnerability in open-webui/open-webui v0.3.8 allows an attacker to view admin details. The application does not verify whether the attacker is an administrator, allowing the attacker to directly call the /api/v1/auths/admin/details interface to retrieve the first admin (owner) details.
Open WebUI Vulnerable to a Session Fixation Attack
A vulnerability in open-webui/open-webui version 0.3.8 allows an attacker with a user-level account to perform a session fixation attack. The session cookie for all users is set with the default `SameSite=Lax` and does not have the `Secure` flag enabled, allowing the session cookie to be sent over HTTP to a cross-origin domain. An attacker can exploit this by embedding a malicious markdown image in a chat, which, when viewed by an administrator, sends the admin's session cookie to the attacker's server. This can lead to a stealthy administrator account takeover, potentially resulting in remote code execution (RCE) due to the elevated privileg
Open WebUI Uncontrolled Resource Consumption vulnerability
A vulnerability in open-webui/open-webui v0.3.8 allows an unauthenticated attacker to sign up with excessively large text in the 'name' field, causing the Admin panel to become unresponsive. This prevents administrators from performing essential user management actions such as deleting, editing, or adding users. The vulnerability can also be exploited by authenticated users with low privileges, leading to the same unresponsive state in the Admin panel.
Open WebUI Uncontrolled Resource Consumption vulnerability
In version v0.3.32 of open-webui/open-webui, the application allows users to submit large payloads in the email and password fields during the sign-in process due to the lack of character length validation on these inputs. This vulnerability can lead to a Denial of Service (DoS) condition when a user submits excessively large strings, exhausting server resources such as CPU, memory, and disk space, and rendering the service unavailable for legitimate users. This makes the server susceptible to resource exhaustion attacks without requiring authentication.
Open WebUI Uncontrolled Resource Consumption vulnerability
In version 0.3.32 of open-webui/open-webui, the absence of authentication mechanisms allows any unauthenticated attacker to access the `api/v1/utils/code/format` endpoint. If a malicious actor sends a POST request with an excessively high volume of content, the server could become completely unresponsive. This could lead to severe performance issues, causing the server to become unresponsive or experience significant degradation, ultimately resulting in service interruptions for legitimate users.
open-webui allows writing and deleting arbitrary files
In version v0.3.8 of open-webui/open-webui, the endpoint /api/pipelines/upload is vulnerable to arbitrary file write and delete due to unsanitized file.filename concatenation with CACHE_DIR. This vulnerability allows attackers to overwrite and delete system files, potentially leading to remote code execution.
A denial of service (DoS) vulnerability exists in zenml-io/zenml version 0.56.3 due to improper handling of line feed (`\n`) characters in component names. When a low-privileged user adds a component through the API endpoint `api/v1/workspaces/default/components` with a name containing a `\n` character, it leads to uncontrolled resource consumption. This vulnerability results in the inability of users to add new components in certain categories (e.g., 'Image Builder') and to register new stacks through the UI, thereby degrading the user experience and potentially rendering the ZenML Dashboard unusable. The issue does not affect component addi
Header injection possibility since DomainNameValidator accepted newlines in input
An issue was discovered in Django 6.0 before 6.0.7 and 5.2 before 5.2.16.
`DomainNameValidator` does not prohibit newlines in domain names (unless used via a form field, since `CharField` strips newlines). If an application uses values with newlines in an HTTP response, header injection can occur. Django itself is unaffected because `HttpResponse` prohibits newlines in HTTP headers.
Earlier, unsupported Django series (such as 5.0.x, 4.1.x, and 3.2.x) were not evaluated and may also be affected.
Django would like to thank Bence Nagy for reporting this issue.
An issue was discovered in Django 6.0 before 6.0.7 and 5.2 before 5.2.16.
`django.contrib.gis.gdal.GDALRaster` over-reads its in-memory buffer when constructed from a bytes object, which can disclose adjacent memory or cause service degradation via a potential segmentation fault when the `vsi_buffer` property is accessed.
Earlier, unsupported Django series (such as 5.0.x, 4.1.x, and 3.2.x) were not evaluated and may also be affected.
Django would like to thank Bence Nagy for reporting this issue.