CVE-2026-85673
LLaMA-Factory contains a server-side request forgery vulnerability in the OpenAI-compatible API multimodal media URL handler that allows unauthenticated attackers to bypass SSRF validation. The check_ssrf_url guard validates URLs once but requests.get follows redirects and re-resolves DNS without re-validation, enabling attackers to use HTTP redirects or DNS rebinding to access internal addresses and cloud metadata endpoints.
脆弱性の種類
LLaMA-Factory contains a server-side request forgery vulnerability in the OpenAI-compatible API multimodal media URL handler that allows unauthenticated attackers to bypass SSRF validation. The check_ssrf_url guard validates URLs once but requests.get follows redirects and re-resolves DNS without re-validation, enabling attackers to use HTTP redirects or DNS rebinding to access internal addresses and cloud metadata endpoints.
参考情報
- https://github.com/hiyouga/LlamaFactory
- https://github.com/hiyouga/LlamaFactory/blob/v0.9.5/src/llamafactory/api/chat.py
- https://github.com/hiyouga/LlamaFactory/blob/v0.9.5/src/llamafactory/api/common.py
- https://github.com/hiyouga/LlamaFactory/issues/10646
- https://www.vulncheck.com/advisories/llama-factory-ssrf-guard-bypass-via-redirect-and-dns-rebinding
- https://github.com/hiyouga/LlamaFactory/issues/10646