BentoML has Dockerfile Command Injection via system_packages in bentofile.yaml
BentoML is a Python library for building online serving systems optimized for AI apps and model inference. Prior to 1.4.37, the `docker.system_packages` field in `bentofile.yaml` accepts arbitrary strings that are interpolated directly into Dockerfile `RUN` commands without sanitization. Since `system_packages` is semantically a list of OS package names (data), users do not expect values to be interpreted as shell commands. A malicious `bentofile.yaml` achieves arbitrary command execution during `bentoml containerize` / `docker build`. Version 1.4.37 fixes the issue.
OpenHands is Vulnerable to Command Injection through its Git Diff Handler
OpenHands is software for AI-driven development. Starting in version 1.5.0, a Command Injection vulnerability exists in the `get_git_diff()` method at `openhands/runtime/utils/git_handler.py:134`. The `path` parameter from the `/api/conversations/{conversation_id}/git/diff` API endpoint is passed unsanitized to a shell command, allowing authenticated attackers to execute arbitrary commands in the agent sandbox. The user is already allowed to instruct the agent to execute commands, but this bypasses the normal channels. Version 1.5.0 fixes the issue.
pypdf: Possible infinite loop during recovery attempts in DictionaryObject.read_from_stream
pypdf is a free and open-source pure-python PDF library. Versions prior to 6.9.2 have a vulnerability in which an attacker can craft a PDF which leads to an infinite loop. This requires reading a file in non-strict mode. This has been fixed in pypdf 6.9.2. If users cannot upgrade yet, consider applying the changes from the patch manually.
vLLM's hardcoded trust_remote_code=True in NemotronVL and KimiK25 bypasses user security opt-out
vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode `trust_remote_code=True` when loading sub-components, bypassing the user's explicit `--trust-remote-code=False` security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust. Version 0.18.0 patches the issue.
Open WebUI's Insecure Direct Object Reference (IDOR) allows access to other users' memories
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to version 0.8.6, any authenticated user can read other users' private memories via `/api/v1/retrieval/query/collection`. Version 0.8.6 patches the issue.
Open WebUI has unauthorized deletion of knowledge files
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to version 0.8.6, an access control check is missing when deleting a file from a knowledge base. The only check being done is that the user has write access to the knowledge base (or is admin), but NOT that the file actually belongs to this knowledge base. It is thus possible to delete arbitrary files from arbitrary knowledge bases (as long as one knows the file id). Version 0.8.6 patches the issue.
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to version 0.8.6, any authenticated user can overwrite any file's content by ID through the `POST /api/v1/retrieval/process/files/batch` endpoint. The endpoint performs no ownership check, so a regular user with read access to a shared knowledge base can obtain file UUIDs via `GET /api/v1/knowledge/{id}/files` and then overwrite those files, escalating from read to write. The overwritten content is served to the LLM via RAG, meaning the attacker controls what the model tells other users. Version 0.8.6 patches the issue.
Open WebUI vulnerable to Path Traversal in `POST /api/v1/audio/transcriptions`
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to version 0.8.6, an unsanitized filename field in the speech-to-text transcription endpoint allows any authenticated non-admin user to trigger a `FileNotFoundError` whose message — including the server's absolute `DATA_DIR` path — is returned verbatim in the HTTP 400 response body, confirming information disclosure on all default deployments. Version 0.8.6 patches the issue.
Streamlit on Windows has Unauthenticated SSRF Vulnerability (NTLM Credential Exposure)
Streamlit is a data oriented application development framework for python. Streamlit Open Source versions prior to 1.54.0 running on Windows hosts have an unauthenticated Server-Side Request Forgery (SSRF) vulnerability. The vulnerability arises from improper validation of attacker-supplied filesystem paths. In certain code paths, including within the `ComponentRequestHandler`, filesystem paths are resolved using `os.path.realpath()` or `Path.resolve()` before sufficient validation occurs. On Windows systems, supplying a malicious UNC path (e.g., `\\attacker-controlled-host\share`) can cause the Streamlit server to initiate outbound SMB conne
MobSF has SQL Injection in its SQLite Database Viewer Utils
MobSF is a mobile application security testing tool used. Prior to version 4.4.6, MobSF's `read_sqlite()` function in `mobsf/MobSF/utils.py` (lines 542-566) uses Python string formatting (`%`) to construct SQL queries with table names read from a SQLite database's `sqlite_master` table. When a security analyst uses MobSF to analyze a malicious mobile application containing a crafted SQLite database, attacker-controlled table names are interpolated directly into SQL queries without parameterization or escaping. This allows an attacker to cause denial of service and achieve SQL injection. Version 4.4.6 patches the issue.
Briefcase: Windows MSI Installer Privilege Escalation via Insecure Directory Permissions
Briefcase is a tool for converting a Python project into a standalone native application. Starting in version 0.3.0 and prior to version 0.3.26, if a developer uses Briefcase to produce an Windows MSI installer for a project, and that project is installed for All Users (i.e., per-machine scope), the installation process creates an directory that inherits all the permissions of the parent directory. Depending on the location chosen by the installing user, this may allow a low privilege but authenticated user to replace or modify the binaries installed by the application. If an administrator then runs the altered binary, the binary will run wit
Modoboa is a mail hosting and management platform. Prior to version 2.7.1, `exec_cmd()` in `modoboa/lib/sysutils.py` always runs subprocess calls with `shell=True`. Since domain names flow directly into shell command strings without any sanitization, a Reseller or SuperAdmin can include shell metacharacters in a domain name to run arbitrary OS commands on the server. Version 2.7.1 patches the issue.
Requests has Insecure Temp File Reuse in its extract_zipped_paths() utility function
Requests is a HTTP library. Prior to version 2.33.0, the `requests.utils.extract_zipped_paths()` utility function uses a predictable filename when extracting files from zip archives into the system temporary directory. If the target file already exists, it is reused without validation. A local attacker with write access to the temp directory could pre-create a malicious file that would be loaded in place of the legitimate one. Standard usage of the Requests library is not affected by this vulnerability. Only applications that call `extract_zipped_paths()` directly are impacted. Starting in version 2.33.0, the library extracts files to a non-d
Two LiteLLM versions published containing credential harvesting malware
After an API Token exposure from an exploited trivy dependency, two new releases of `litellm` were uploaded to PyPI containing automatically activated malware, harvesting sensitive credentials and files, and exfiltrating to a remote API.
Anyone who has installed and run the project should assume any credentials available to litellm environment may have been exposed, and revoke/rotate thema ccordingly.
NVIDIA NeMo Framework contains a vulnerability where an attacker may cause remote code execution. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, information disclosure and data tampering.
NVIDIA NeMo Framework contains a vulnerability in checkpoint loading where an attacker could cause remote code execution. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, information disclosure and data tampering.
NiceGUI's unvalidated chunk size parameter in media routes can cause memory exhaustion
NiceGUI is a Python-based UI framework. Prior to version 3.9.0, NiceGUI's app.add_media_file() and app.add_media_files() media routes accept a user-controlled query parameter that influences how files are read during streaming. The parameter is passed to the range-response implementation without validation, allowing an attacker to bypass chunked streaming and force the server to load entire files into memory at once. With large media files and concurrent requests, this can lead to excessive memory consumption, degraded performance, or denial of service. This issue has been patched in version 3.9.0.
pyload-ng: SETTINGS Permission Users Can Achieve Remote Code Execution via Unrestricted Reconnect Script Configuration
pyLoad is a free and open-source download manager written in Python. From version 0.4.0 to before version 0.5.0b3.dev97, the set_config_value() API endpoint allows users with the non-admin SETTINGS permission to modify any configuration option without restriction. The reconnect.script config option controls a file path that is passed directly to subprocess.run() in the thread manager's reconnect logic. A SETTINGS user can set this to any executable file on the system, achieving Remote Code Execution. The only validation in set_config_value() is a hardcoded check for general.storage_folder — all other security-critical settings including recon