high
ModeLeak: LLM Model Exfiltration Vulnerability in Vertex AI
Published Tue, Nov 12th, 2024
Platforms
A vulnerability in GCP's Vertex AI service allows privilege escalation and unauthorized access to sensitive LLM models. Attackers can exfiltrate these models by exploiting misconfigurations in access controls and service bindings. By exploiting custom job permissions, researchers were able to escalate their privileges and gain unauthorized access to all data services in the project. In addition, deploying a poisoned model in Vertex AI led to the exfiltration of all other fine-tuned models, posing a proprietary and sensitive data exfiltration attack risk.
VertexAI
None required
No tracked CVEs
Contributed by https://github.com/OfirBalassiano
Entry Status
Finalized
Disclosure Date
-
Exploitability Period
Known ITW Exploitation
Detection Methods
None
Piercing Index Rating
Discovered by
Ofir Balassiano, Ofir Shaty, Palo Alto Networks
high
Multiple vulnerabilities were discovered in Google's Cloud Architecture Diagramming Tool, including XSS, unauthorized access to user data, and misconfigured storage buckets. The issues allowed acce...
Multiple security vulnerabilities were identified in data.all, an open source development framework for building data marketplaces on AWS. The issues affect versions 1.0.0 through 2.6.0 and include...
A technique called "repo swatting" allows attackers to delete GitHub and block GitLab accounts by exploiting file upload features and abuse reporting mechanisms. Attackers upload malicious files to...