Policy bypass via a client-controlled X-Original-URI header
A client-controlled header was trusted before the request path, allowing matching ALLOW rules to bypass the Anubis challenge.
vulnerability research
cves
Published vulnerabilities from my security research.
Additional findings are moving through coordinated disclosure.
- published CVEs
- 13
- projects
- 06
- critical
- 04
Anubis
Web AI firewall · 1 CVE
ChromaDB
AI-native database · 6 CVEs
-
Pre-authentication remote code execution
An unauthenticated attacker could submit a malicious model repository with trust_remote_code enabled and execute code on the server.
-
Authenticated remote code execution
A user with collection-update permission could load a malicious model repository with trust_remote_code and execute code on the server.
-
Cross-tenant authorization bypass in ChromaDB Rust
Any authenticated user could read, write, update, or delete collections belonging to another tenant.
-
Cross-tenant authorization bypass in ChromaDB Python
Any authenticated user could read, write, update, or delete collections belonging to another tenant.
-
SimpleRBAC cross-tenant authorization bypass
The provider checked whether a user held a permission, but not which tenant, database, or collection that permission applied to.
-
Authorization bypass through V1 collection endpoints
The V1 endpoints passed no tenant or database to the authorization layer, allowing its controls to be bypassed.
Flair
NLP framework · 1 CVE
-
Arbitrary code execution through model deserialization
Loading a malicious language model could trigger unsafe deserialization and execute arbitrary code.
MISP
Threat intelligence platform · 1 CVE
-
Path traversal in the EventReport image viewer
A site administrator could traverse outside the intended path when viewing an EventReport picture.
Keras
Deep-learning framework · 1 CVE
-
Safe-mode bypass through TorchModuleWrapper
A malicious Keras file could execute arbitrary code when loaded, despite safe mode being enabled.
Backend.AI
AI compute platform · 3 CVEs
-
Missing authorization for interactive sessions
An attacker could take over active sessions and access, steal, or alter data available inside them.
-
Arbitrary account creation through missing access control
Unauthenticated users could create accounts and reach private data even when registration was disabled.
-
Credential exposure leading to account takeover
Sensitive data exposed through active sessions could reveal management-platform credentials.
Scores shown are CNA-provided base scores from the linked CVE records.