Audit diffusion models
before they reach production.
Membership inference attack platform for privacy-aware AI compliance.
Coverage
Membership inference risk, three layers of audit depth.
01
Black-box signals
Start with the fastest read on exposure so teams can screen risk early and brief stakeholders clearly.
02
Gray-box and white-box depth
When the story needs stronger evidence, keep drilling into training traces, attack effects, and comparisons.
03
One workspace
The homepage frames the system. The workspace carries tasks, runtime status, and exports.
Flow
The homepage sets scope. The workspace handles execution.
The homepage covers scope and entry points. Tasks, runtime status, and report exports live in the workspace.
A
Sign in
Authenticate through one entry point and stay inside the same site.
B
Create a job
Open the audit flow, choose the target, launch the run, and follow its state.
C
Review results
Return to the workspace and report pages to collect metrics and package the outcome.
Get started