A vulnerability, which was classified as problematic, was found in Zylon PrivateGPT up to 0.6.2. This affects an unknown part of the file settings.yaml. The manipulation of the argument allow_origins leads to permissive cross-domain policy with untrusted domains. It is possible to initiate the attack remotely. The exploit has been disclosed to the public and may be used. The vendor was contacted early about this disclosure but did not respond in any way.
A Python command injection vulnerability exists in the `SagemakerLLM` class's `complete()` method within `./private_gpt/components/llm/custom/sagemaker.py` of the imartinez/privategpt application, versions up to and including 0.3.0. The vulnerability arises due to the use of the `eval()` function to parse a string received from a remote AWS SageMaker LLM endpoint into a dictionary. This method of parsing is unsafe as it can execute arbitrary Python code contained within the response. An attacker can exploit this vulnerability by manipulating the response from the AWS SageMaker LLM endpoint to include malicious Python code, leading to potential execution of arbitrary commands on the system hosting the application. The issue is fixed in version 0.6.0.
imartinez/privategpt version 0.2.0 is vulnerable to a local file inclusion vulnerability that allows attackers to read arbitrary files from the filesystem. By manipulating file upload functionality to ingest arbitrary local files, attackers can exploit the 'Search in Docs' feature or query the AI to retrieve or disclose the contents of any file on the system. This vulnerability could lead to various impacts, including but not limited to remote code execution by obtaining private SSH keys, unauthorized access to private files, source code disclosure facilitating further attacks, and exposure of configuration files.
A stored Cross-Site Scripting (XSS) vulnerability exists in the 'imartinez/privategpt' repository due to improper validation of file uploads. Attackers can exploit this vulnerability by uploading malicious HTML files, such as those containing JavaScript payloads, which are then executed in the context of the victim's session when accessed. This could lead to the execution of arbitrary JavaScript code in the context of the user's browser session, potentially resulting in phishing attacks or other malicious actions. The vulnerability affects the latest version of the repository.