https://vulnerability.circl.lu Contains only the most 10 recent entries. http://www.rssboard.org/rss-specification python-feedgen en Sun, 04 Oct 2026 13:14:24 +0000 https://vulnerability.circl.lu/vuln/cve-2025-53002 <p><strong>Affected:</strong> hiyouga LLaMA-Factory</p> <p>LLaMA-Factory is a tuning library for large language models. A remote code execution vulnerability was discovered in LLaMA-Factory versions up to and including 0.9.3 during the LLaMA-Factory training process. This vulnerability arises because the `vhead_file` is loaded without proper safeguards, allowing malicious attackers to execute arbitrary malicious code on the host system simply by passing a malicious `Checkpoint path` parameter through the `WebUI` interface. The attack is stealthy, as the victim remains unaware of the exploitation. The root cause is that the `vhead_file` argument is loaded without the secure parameter `weights_only=True`. Version 0.9.4 contains a fix for the issue.</p> <p><strong>Affected:</strong> hiyouga LLaMA-Factory</p> <p>LLaMA-Factory is a tuning library for large language models. A remote code execution vulnerability was discovered in LLaMA-Factory versions up to and including 0.9.3 during the LLaMA-Factory training process. This vulnerability arises because the `vhead_file` is loaded without proper safeguards, allowing malicious attackers to execute arbitrary malicious code on the host system simply by passing a malicious `Checkpoint path` parameter through the `WebUI` interface. The attack is stealthy, as the victim remains unaware of the exploitation. The root cause is that the `vhead_file` argument is loaded without the secure parameter `weights_only=True`. Version 0.9.4 contains a fix for the issue.</p> https://vulnerability.circl.lu/vuln/cve-2025-53002 https://vulnerability.circl.lu/vuln/ghsa-xj56-p8mm-qmxj <p><strong>Affected:</strong> PyPI: llamafactory</p> <p>### Summary A critical remote code execution vulnerability was discovered during the Llama Factory training process. This vulnerability arises because the `vhead_file` is loaded without proper safeguards, allowing malicious attackers to execute arbitrary malicious code on the host system simply by passing a malicious `Checkpoint path` parameter through the `WebUI` interface. The attack is stealthy, as the victim remains unaware of the exploitation. The root cause is that the `vhead_file` argument is loaded without the secure parameter `weights_only=True`.</p> <p>Note: In torch versions &lt;2.6, the default setting is `weights_only=False`, and Llama Factory&#39;s `setup.py` only requires `torch&gt;=2.0.0`.</p> <p>### Affected Version</p> <p>Llama Factory versions &lt;=0.9.3 are affected by this vulnerability.</p> <p>### Details</p> <p>1. In LLaMA Factory&#39;s WebUI, when a user sets the `Checkpoint path`, it modifies the `adapter_name_or_path` parameter passed to the training process. code in src/llamafactory/webui/runner.py &lt;img width=&#34;1040&#34; alt=&#34;image-1&#34; src=&#34;https://github.com/user-attachments/assets/c8bc79e4-ce7d-43c9-b0fd-e37c235e6585&#34; /&gt;</p> <p>2. The `adapter_name_or_path` passed to the training process is then used in `src/llamafactory/model/model_utils/valuehead.py` to fetch the corresponding `value_head.bin` file from Hugging Face. This file is subsequently loaded via `torch.load()` without the security parameter `weights_only=True` being set, resulting in remote code execution. code in src/llamafactory/model/model_u…</p> <p><strong>Affected:</strong> PyPI: llamafactory</p> <p>### Summary A critical remote code execution vulnerability was discovered during the Llama Factory training process. This vulnerability arises because the `vhead_file` is loaded without proper safeguards, allowing malicious attackers to execute arbitrary malicious code on the host system simply by passing a malicious `Checkpoint path` parameter through the `WebUI` interface. The attack is stealthy, as the victim remains unaware of the exploitation. The root cause is that the `vhead_file` argument is loaded without the secure parameter `weights_only=True`.</p> <p>Note: In torch versions &lt;2.6, the default setting is `weights_only=False`, and Llama Factory&#39;s `setup.py` only requires `torch&gt;=2.0.0`.</p> <p>### Affected Version</p> <p>Llama Factory versions &lt;=0.9.3 are affected by this vulnerability.</p> <p>### Details</p> <p>1. In LLaMA Factory&#39;s WebUI, when a user sets the `Checkpoint path`, it modifies the `adapter_name_or_path` parameter passed to the training process. code in src/llamafactory/webui/runner.py &lt;img width=&#34;1040&#34; alt=&#34;image-1&#34; src=&#34;https://github.com/user-attachments/assets/c8bc79e4-ce7d-43c9-b0fd-e37c235e6585&#34; /&gt;</p> <p>2. The `adapter_name_or_path` passed to the training process is then used in `src/llamafactory/model/model_utils/valuehead.py` to fetch the corresponding `value_head.bin` file from Hugging Face. This file is subsequently loaded via `torch.load()` without the security parameter `weights_only=True` being set, resulting in remote code execution. code in src/llamafactory/model/model_u…</p> https://vulnerability.circl.lu/vuln/ghsa-xj56-p8mm-qmxj