Vbcrlf

An issue exists in components based on `BaseFileComponent`, including Docling (`DoclingInlineComponent`), Docling Serve (`DoclingRemoteComponent`), Read File (`FileComponent`), NVIDIA Retriever Extraction (`NvidiaIngestComponent`), Video File (`VideoFileComponent`), and Unstructured API (`UnstructuredComponent`). The ` unpack bundle()` function in `langflow/src/lfx/src/lfx/base/data/base file.py` fails to properly validate files during TAR extraction, allowing the processing of symlinks. An attacker can provide a crafted TAR file containing symlinks to absolute paths on the file-system, directing the node to read arbitrary files. This can lead to the exposure of sensitive information, such as the `secret key` used for JWT tokens, which could subsequently allow an attacker to bypass authentication and achieve remote code execution by creating a flow with a Python Interpreter node.

Unauthenticated users with network access can upload unlimited amounts of data to the server, which can lead to disk space exhaustion and a resulting denial-of-service. Additionally, the server response discloses the absolute path of the uploaded file, creating an information leak that could be used to facilitate further attacks. This issue occurs at the '/api/v1/upload/{flow id}' endpoint via the `create upload file()` function, where there is a lack of authentication and validation for the `flow id` variable. Over 38,300 potentially affected instances were identified via FOFA.

The Shareable Playground feature, also known as Public Flows, allows unauthenticated users to execute workflows via a public link. A flaw in this feature enables arbitrary file reading depending on the flow configuration. Specifically, the endpoint '/api/v1/build public tmp' accepts a `files` parameter containing a list of file paths. These paths can be local files or S3 paths, which are then read by the `to lc message` function within `LCModelComponent. get chat result` and fed into the Large Language Model (LLM) as images. The ability to retrieve the file contents depends on the specific LLM configuration used.

The Shareable Playground (also known as Public Flows) feature allows unauthenticated users to execute workflows via a link. A flaw in the '/api/v1/build public tmp' endpoint enables attackers to execute arbitrary Python code on the server by submitting a crafted JSON payload. The issue occurs because the system allows providing custom Python code within the `data.nodes[X].data.node.template.code.value` variable during flow execution. Over 30,900 potentially affected deployments were identified via FOFA. This can lead to full system compromise, data theft, or service disruption.