ms-swift 4.5.2 Unauthenticated SSRF via Multimodal Media URLs

ms-swift 4.5.2 contains a server-side request forgery vulnerability in the swift deploy OpenAI-compatible API that fetches multimodal media URLs without validation or redirect filtering. Unauthenticated attackers can supply arbitrary image_url, audio_url, or video_url parameters to make the server issue requests to internal services and cloud metadata endpoints.

ModelScope through 1.40.0 Unsafe YAML Deserialization in Model Config Loading

ModelScope uses PyYAML's unsafe yaml.Loader to parse model configuration files, allowing arbitrary code execution through Python object construction tags. Attackers can craft malicious model repositories with poisoned configuration files that execute code when loaded by users.

modelscope ms-swift PIL Image Cache Key base.py Template._save_pil_image weak hash

A security vulnerability has been detected in modelscope ms-swift up to 4.2.0. This affects the function Template._save_pil_image of the file swift/template/base.py of the component PIL Image Cache Key Handler. The manipulation leads to use of weak hash. An attack has to be approached locally. A high degree of complexity is needed for the attack. It is indicated that the exploitability is difficult. The exploit has been disclosed publicly and may be used. The pull request to fix this issue awaits acceptance.

ModelScope: ModelScope: Arbitrary code execution via crafted configuration module

An issue was discovered in ModelScope 1.25.0 allowing attackers to execute arbitrary code via crafted module listed in the configuration file (dey_mini.yaml) under the key ['nnet']['module'].

modelscope agentscope _python.py execute_shell_command code injection

A vulnerability was determined in modelscope agentscope up to 1.0.18. Affected by this vulnerability is the function execute_python_code/execute_shell_command of the file src/AgentScope/tool/_coding/_python.py. This manipulation causes code injection. The attack is possible to be carried out remotely. The exploit has been publicly disclosed and may be utilized. The vendor was contacted early about this disclosure but did not respond in any way.

modelscope agentscope _agent_base.py _process_audio_block server-side request forgery

A weakness has been identified in modelscope agentscope up to 1.0.18. This vulnerability affects the function _process_audio_block of the file src/agentscope/agent/_agent_base.py. Executing a manipulation of the argument url can lead to server-side request forgery. It is possible to launch the attack remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

modelscope agentscope Internal Service _common.py _get_bytes_from_web_url server-side request forgery

A security flaw has been discovered in modelscope agentscope up to 1.0.18. This affects the function _get_bytes_from_web_url of the file src/agentscope/_utils/_common.py of the component Internal Service. Performing a manipulation results in server-side request forgery. It is possible to initiate the attack remotely. The exploit has been released to the public and may be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

modelscope agentscope Cloud Metadata Endpoint _openai_tools.py openai_audio_to_text server-side request forgery

A vulnerability was identified in modelscope agentscope up to 1.0.18. Affected by this issue is the function _parse_url/prepare_image/openai_audio_to_text of the file src/agentscope/tool/_multi_modality/_openai_tools.py of the component Cloud Metadata Endpoint. Such manipulation of the argument image_url/audio_file_url leads to server-side request forgery. The attack may be performed from remote. The exploit is publicly available and might be used. The vendor was contacted early about this disclosure but did not respond in any way.

Command injection vulnerability in ModelScope's ms-agent

A command injection vulnerability in ModelScope's ms-agent versions v1.6.0rc1 and earlier exists, allowing an attacker to execute arbitrary operating system commands through crafted prompt-derived input.

A remote code execution (RCE) vulnerability exists in the ms-swift project version 3.3.0 due to unsafe deserialization in tests/run.py using yaml.load() from the PyYAML library (versions = 5.3.1). If an attacker can control the content of the YAML configuration file passed to the --run_config parameter, arbitrary code can be executed during deserialization. This can lead to full system compromise. The vulnerability is triggered when a malicious YAML file is loaded, allowing the execution of arbitrary Python commands such as os.system(). It is recommended to upgrade PyYAML to version 5.4 or higher, and to use yaml.safe_load() to mitigate the issue.

CSRF due to overly permissive CORS headers in modelscope/agentscope

A vulnerability in modelscope/agentscope, specifically in the AgentScope Studio backend server, allows for Cross-Site Request Forgery (CSRF) due to overly permissive CORS headers. This issue affects the latest commit on the main branch (21161fe). The vulnerability permits an attacker to access all backend endpoints, including the `api/file` endpoint, enabling the reading of arbitrary files on the target's local file system through CSRF.

Remote Code Execution via Deserialization in modelscope/agentscope

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.

The modelscope/ms-swift library thru 2.6.1 is vulnerable to arbitrary code execution through deserialization of untrusted data within the `load_model_meta()` function of the `ModelFileSystemCache()` class. Attackers can execute arbitrary code and commands by crafting a malicious serialized `.mdl` payload, exploiting the use of `pickle.load()` on data from potentially untrusted sources. This vulnerability allows for remote code execution (RCE) by deceiving victims into loading a seemingly harmless checkpoint during a normal training process, thereby enabling attackers to execute arbitrary code on the targeted machine. Note that the payload file is a hidden file, making it difficult for the victim to detect tampering. More importantly, during the model training process, after the `.mdl` file is loaded and executes arbitrary code, the normal training process remains unaffected'meaning the user remains unaware of the arbitrary code execution.

Directory Traversal in modelscope/agentscope

A directory traversal vulnerability exists in modelscope/agentscope version 0.0.4. An attacker can exploit this vulnerability to read any local JSON file by sending a crafted POST request to the /read-examples endpoint.

Path Traversal in modelscope/agentscope

A path traversal vulnerability exists in the save-workflow and load-workflow functionality of modelscope/agentscope versions prior to the fix. This vulnerability allows an attacker to read and write arbitrary JSON files on the filesystem, potentially leading to the exposure or modification of sensitive information such as configuration files, API keys, and hardcoded passwords.

In agentscope <=v0.0.4, the file agentscope\web\workstation\workflow_utils.py has the function is_callable_expression. Within this function, the line result = eval(s) poses a security risk as it can directly execute user-provided commands.

Path Traversal in modelscope/agentscope

A path traversal vulnerability exists in the modelscope/agentscope application, affecting all versions. The vulnerability is present in the /delete-workflow endpoint, allowing an attacker to delete arbitrary files from the filesystem. This issue arises due to improper input validation, enabling the attacker to manipulate file paths and delete sensitive files outside of the intended directory.

Path Traversal in modelscope/agentscope

A path traversal vulnerability exists in modelscope/agentscope version v.0.0.4. The API endpoint `/api/file` does not properly sanitize the `path` parameter, allowing an attacker to read arbitrary files on the server.

Local File Inclusion (LFI) in modelscope/agentscope

A Local File Inclusion (LFI) vulnerability exists in the /load-workflow endpoint of modelscope/agentscope version v0.0.4. This vulnerability allows an attacker to read arbitrary files from the server, including sensitive files such as API keys, by manipulating the filename parameter. The issue arises due to improper sanitization of user input passed to the os.path.join function, which can be exploited to access files outside the intended directory.

CORS Vulnerability in modelscope/agentscope

A Cross-Origin Resource Sharing (CORS) vulnerability exists in modelscope/agentscope version v0.0.4. The CORS configuration on the agentscope server does not properly restrict access to only trusted origins, allowing any external domain to make requests to the API. This can lead to unauthorized data access, information disclosure, and potential further exploitation, thereby compromising the integrity and confidentiality of the system.