CVE-2026-79785: X-AnyLabeling before 4.0.0-beta.9 Improper Certificate Validation in Model Downloads
X-AnyLabeling's model downloader disabled TLS certificate verification. downloadwithretry in anylabeling/services/autolabeling/model.py built a context with ssl.createunverifiedcontext() and passed it to urllib.request.urlopen, so neither the certificate chain nor the hostname was checked on any model download, and models are fetched over HTTPS from the project's release host. Any party positioned to intercept that connection could therefore answer it with content of their own choosing. The response is written to a .part file and moved into place with os.replace, and the only post-download check, safecheckmodel, validates the file's format rather than its provenance: no hash or signature is compared against an expected value. For an ONNX target the substituted file passes onnx.checker.checkmodel and is then used for inference, so the attacker chooses the model that produces the application's annotations. For a .pth or .pt target, which the shipped SAM2 video, YOLOE, UPN and openvision configurations use, the check worker calls torch.load without weightsonly, so a substituted file is unpickled and executes code of the attacker's choosing on PyTorch releases predating the weightsonly default.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
X-AnyLabelingto a version that resolves this vulnerability.Fixed in 4.0.0-beta.9Patch X-AnyLabeling before 4.0.0-beta.9 Improper Certificate Validation in Model Downloads - Configuration
Remove use of ssl._create_unverified_context() in download_with_retry and use a normal TLS context so urllib.request.urlopen validates the certificate chain and hostname for HTTPS model downloads.
X-AnyLabeling model downloader (anylabeling/services/auto_labeling/model.py) ssl._create_unverified_context() usage in urllib.request.urlopen TLS context = enabled TLS certificate verification
Event History
Frequently Asked Questions
Who is exposed to this issue?
Users of X-AnyLabeling versions before 4.0.0-beta.9 are exposed when the application downloads models from the project's release host. An attacker must be able to intercept or otherwise control the HTTPS connection used for that download.
What can an attacker do with a substituted model?
A substituted ONNX model that passes format validation can control the annotations produced during inference. For .pth or .pt models used by the shipped SAM2 video, YOLOE, UPN, and open_vision configurations, loading the attacker-supplied file can execute attacker-chosen code on affected PyTorch releases.
Are downloaded models verified for authenticity before use?
No. The download path did not validate the server certificate or hostname, and it did not compare the downloaded file against an expected hash or signature. The post-download check validates file format rather than provenance.
What version addresses the issue?
The issue affects versions before 4.0.0-beta.9. The referenced fix is associated with the v4.0.0-beta.9 release.