AIMS SECURE TESTING ENVIRONMENTDETERMINISTIC SCORING

AIMS Module Testing Console

Continuous adversarial testing for LLM-backed applications: prompt injection, data leakage, model inversion, and hallucination risk, mapped to the OWASP LLM Top 10.

Overview

Posture dashboard

No runs recorded yet in this browser.

Modules

Testing coverage

Seven independently scoped modules, each mapped to a risk category.

T-01LLM01

Prompt Injection

Jailbreak / instruction-override attempts against the system prompt.

Not yet run

T-02LLM06

Prompt Leakage

Deterministic canary-token test for system-prompt disclosure.

Not yet run

T-03LLM04

Resource Limits

Single size-capped request checking for a denial-of-service guardrail.

Not yet run

T-04LLM06

Log Analysis

Passive regex scan (not a live query) of a pasted or uploaded transcript for secret, PII, and injection-attempt patterns.

Not yet run

T-05LLM06

PII / SII Exposure

Seeded fake customer record; flags verbatim disclosure of protected fields.

Not yet run

T-06PRIVACY

Model Inversion

Tests whether redacted fields can be reconstructed from partial context.

Not yet run

T-07LLM09

Hallucination

Invented, nonexistent entities. Scores confident fabrication vs. honest uncertainty.

Not yet run