CVE-2024-35198: TorchServe allowed_urls check bypass that permits unvalidated model downloads

Summary

TorchServe's allowed_urls check can be bypassed when a URL contains characters such as "..", and the model is still downloaded into the model store. A downloaded file can then be referenced without a URL on later requests, which defeats the allowed_urls security check. Customers using PyTorch inference Deep Learning Containers through Amazon SageMaker and EKS are not affected.

Mitigation

This issue in TorchServe has been fixed by validating the URL without characters such as ".." before downloading, see PR #3082. TorchServe release 0.11.0 includes the fix. Users are advised to upgrade.

Affected packages in the Exposure Registry

Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.

• torchservePyPINo LLM dependency detected

Related items

• MediumCVE-2026-121936: easyVoice server-side request forgery through openaiBaseUrl in /api/v1/ttsSimilar attack · NVD/CVE Database
• MediumCVE-2026-118628: MLflow get-artifacts server-side request forgery information disclosureSimilar attack · NVD/CVE Database
• HighCVE-2026-118625: MLflow Tracking Server privilege escalation via get_endpointsSimilar attack · NVD/CVE Database
• InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSimilar attack · Elastic Security Labs
• MediumCritical Langflow flaws enable authentication bypass and RCESimilar attack · CSO Online