#vulnerability

A vulnerability in the RpcAgentServerLauncher class of modelscope/agentscope v0.0.6a3 allows for remote code execution (RCE) via deserialization of untrusted data using the dill library. The issue occurs in the AgentServerServicer.create_agent method, where serialized input is deserialized using dill.loads, enabling an attacker to execute arbitrary commands on the server.

In version 3.22.0 of aimhubio/aim, the AimQL query language uses an outdated version of the safer_getattr() function from RestrictedPython. This version does not protect against the str.format_map() method, allowing an attacker to leak server-side secrets or potentially gain unrestricted code execution. The vulnerability arises because str.format_map() can read arbitrary attributes of Python objects, enabling attackers to access sensitive variables such as os.environ. If an attacker can write files to a known location on the Aim server, they can use str.format_map() to load a malicious .dll/.so file into the Python interpreter, leading to unrestricted code execution.

A CORS (Cross-Origin Resource Sharing) misconfiguration in prefecthq/prefect prior ro version 3.0.3 allows unauthorized domains to access sensitive data. This vulnerability can lead to unauthorized access to the database, resulting in potential data leaks, loss of confidentiality, service disruption, and data integrity risks.

A vulnerability in the typeahead endpoint of h2oai/h2o-3 version 3.46.0 allows for a denial of service. The endpoint performs a `HEAD` request to verify the existence of a specified resource without setting a timeout. An attacker can exploit this by sending multiple requests to an attacker-controlled server that hangs, causing the application to block and become unresponsive to other requests.

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.

A vulnerability in the `/3/ImportFiles` endpoint of h2oai/h2o-3 version 3.46.1 allows an attacker to cause a denial of service. The endpoint takes a single GET parameter, `path`, which can be recursively set to reference itself. This leads the server to repeatedly call its own endpoint, eventually filling up the request queue and leaving the server unable to handle other requests.

In h2oai/h2o-3 version 3.46.0.2, a vulnerability exists where uploading and repeatedly parsing a large GZIP file can cause a denial of service. The server becomes unresponsive due to memory exhaustion and a large number of concurrent slow-running jobs. This issue arises from the improper handling of highly compressed data, leading to significant data amplification.

BerriAI/litellm version 1.40.12 contains a vulnerability that allows remote code execution. The issue exists in the handling of the 'post_call_rules' configuration, where a callback function can be added. The provided value is split at the final '.' mark, with the last part considered the function name and the remaining part appended with the '.py' extension and imported. This allows an attacker to set a system method, such as 'os.system', as a callback, enabling the execution of arbitrary commands when a chat response is processed.

In h2oai/h2o-3 version 3.46.0, an endpoint exposing a custom EncryptionTool allows an attacker to encrypt any files on the target server with a key of their choosing. The chosen key can also be overwritten, resulting in ransomware-like behavior. This vulnerability makes it possible for an attacker to encrypt arbitrary files with keys of their choice, making it exceedingly difficult for the target to recover the keys needed for decryption.

In h2oai/h2o-3 version 3.46.0, the endpoint for exporting models does not restrict the export location, allowing an attacker to export a model to any file in the server's file structure, thereby overwriting it. This vulnerability can be exploited to overwrite any file on the target server with a trained model file, although the content of the overwrite is not controllable by the attacker.