CVE-2026-68770
sentence-transformers contains a security control bypass vulnerability that allows attackers to achieve arbitrary code execution by exploiting a logic flaw in the import_module_class helper within sentence_transformers/util/misc.py, where the guard condition includes an 'or os.path.exists(model_name_or_path)' clause that satisfies the trust gate whenever the supplied path exists on the local filesystem, regardless of the trust_remote_code=False argument. Attackers who can control or influence the contents of a model directory on disk can place malicious Python files such as modeling_*.py referenced via modules.json, causing the code to execute at import time when an application loads the model with SentenceTransformer(path, trust_remote_code=False), bypassing the documented security contract and achieving code execution within the loading process.
Schwachstellenklasse
sentence-transformers contains a security control bypass vulnerability that allows attackers to achieve arbitrary code execution by exploiting a logic flaw in the import_module_class helper within sentence_transformers/util/misc.py, where the guard condition includes an 'or os.path.exists(model_name_or_path)' clause that satisfies the trust gate whenever the supplied path exists on the local filesystem, regardless of the trust_remote_code=False argument. Attackers who can control or influence the contents of a model directory on disk can place malicious Python files such as modeling_*.py referenced via modules.json, causing the code to execute at import time when an application loads the model with SentenceTransformer(path, trust_remote_code=False), bypassing the documented security contract and achieving code execution within the loading process.
Quellen
- https://github.com/huggingface/sentence-transformers
- https://github.com/huggingface/sentence-transformers/commit/ae1acc3fb2aa2004577b297eb4a915ce7a03316a
- https://github.com/huggingface/sentence-transformers/issues/3801
- https://github.com/huggingface/sentence-transformers/pull/3807
- https://www.vulncheck.com/advisories/sentence-transformers-arbitrary-code-execution-on-local-model-load-despite-trust-remote-code-false
- https://github.com/huggingface/sentence-transformers/issues/3801
Finden Sie die Schwachstelle, bevor es ein Angreifer tut.
Melden Sie sich mit GitHub an und starten Sie Ihr erstes Audit in weniger als einer Minute. Für den kostenlosen Plan ist keine Kreditkarte nötig.