Skip to content

Security: arm/ai-ml-sdk-vgf-library

Security

SECURITY.md

Security Policy

This software is verified for security for official releases and as such does not make promises about the quality of the product for patches delivered between releases.

Trust model for VGF inputs

VGF files (including embedded SPIR-V™) are expected to be provided by trusted tooling in the build and deployment pipeline. The decoder assumes VGF content is trusted input and does not attempt to sanitize or sandbox SPIR-V™ payloads. Safeguards to improve robustness are in place, but they do not change the trust boundary. Consumers that accept VGF from untrusted sources must implement their own validation/sandboxing at the application boundary. Callers must perform VGF header decoding to check section offsets and sizes against the file length before any decoder is constructed.

Reporting a Vulnerability

Security vulnerabilities may be reported to the Arm® Product Security Incident Response Team (PSIRT) by sending an email to psirt@arm.com.

For more information visit https://developer.arm.com/support/arm-security-updates/report-security-vulnerabilities

There aren't any published security advisories