GHSA-3wp3-xxj9-5jqq: Open WebUI: Cross-user model-list exposure via static cache key in get_all_models (aiocache key= vs key_builder= misuse)

Summary

The `get_all_models` handlers in `routers/openai.py` and `routers/ollama.py` in Open WebUI wrap their permission-filtered model lists in `@cached`, but pass `key=` instead of `key_builder=`. In aiocache 0.12.3, this makes every caller share one static cache entry within `MODELS_CACHE_TTL` (default 1 second), so one user can receive another user's model list, including models they are not permitted to see. The attacker cannot choose the victim or force a target's list into the cache.

Mitigation

Replace `key=` with `key_builder=` at both call sites and make the lambda take the function as its first argument, keying the cache on `user.id`, as in: @cached(ttl=MODELS_CACHE_TTL, key_builder=lambda _func, request, user=None: (f'openai_all_models_{user.id}' if user else 'openai_all_models'))

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.

• open-webuiPyPILLM dependency since 2024-05-20

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