### Impact TorchGeo 0.4–0.6.0 used an [`eval`](https://docs.python.org/3/library/functions.html#eval) statement in its model weight API that could allow an unauthenticated, remote attacker to execute arbitrary commands. All platforms that expose [`torchgeo.models.get_weight()`](https://torchgeo.readthedocs.io/en/v0.6.0/api/models.html#torchgeo.models.get_weight) or [`torchgeo.trainers`](https://torchgeo.readthedocs.io/en/v0.6.0/api/trainers.html) as an external API could be affected. ### Patches The `eval` statement was replaced with a fixed enum lookup, preventing arbitrary code injection. All users are encouraged to upgrade to TorchGeo 0.6.1 or newer. ### Workarounds In unpatched versions, input validation and sanitization can be used to avoid this vulnerability. ### References #### Bug history * Introduced: https://github.com/torchgeo/torchgeo/pull/917 * Patched: https://github.com/torchgeo/torchgeo/pull/2323 * Released: [v0.6.1](https://github.com/microsoft/torchgeo/releases/tag/v0.6.1)
| Vendor | Product | Versions |
|---|---|---|
| pypa | torchgeo | pip/torchgeo: >= 0.4, <= 0.6.0 |
Downstream vendors/products affected by this vulnerability
| Vendor | Product | Source | Confidence |
|---|---|---|---|
| microsoft | apps | cert_advisory | 90% |
