SECENG WORKBENCH

Retrieval & Context Security Test Harness

Test whether your RAG system retrieves what it should — and nothing it should not.

The SecEng RAG Test Harness validates retrieval boundaries, authorization rules, and context integrity across multi-tenant knowledge bases. Detect cross-tenant leaks, poisoned corpus content, and stale-permission access with a structured test suite that maps findings to ISO 42001, NIST AI RMF, and EU AI Act controls.

Explore RAG →See It in Action

WAS RETRIEVAL AUTHORIZED?

Boundary Testing

Cross-tenant, cross-role, stale-permission, and revoked-user test cases.

Source Provenance

Validate citation accuracy and block source laundering.

Privacy Gates

Detect PII, regulated data, and confidential chunks in retrieved context.

Content Validation

Identify poisoned documents and indirect prompt injection in the corpus.

SecEng RAG Test Harness — retrieval pipeline showing policy check, context window, tenant boundary tests, and leakage by type

97.6%

AuthZ pass rate (+2.4% vs last 30 days)

Context leaks detected (1.8% of queries)

Policy violations (1.3% of queries)

124

Tenant boundary tests (108 pass / 12 fail)

Core capabilities

What SecEng RAG Test Harness does.

Multi-Identity Retrieval Testing

Test cross-user, cross-tenant, revoked user, contractor, admin, and external guest access cases. Validate that retrieval respects identity and authorization — not just query relevance scores.

Corpus Seeding & Poison Testing

Seed sensitive documents, poisoned documents, stale-permission files, and indirect prompt-injection content into the corpus. Confirm hostile or unauthorized content is blocked before context.

Policy Check & Context Boundary Validation

Validate that retrieved chunks pass policy checks before entering the context window. Detect the '4 Blocked / 16 Allowed' split — and confirm blocked content never reaches the model response.

Leakage by Type Classification

Classify every leakage event: Cross-Tenant (7), Cross-Role (6), Stale Permission (4), Poisoned Content (3), Source Laundering (2). Prioritize remediation by type and frequency.

Full Pipeline Evidence Capture

Capture query → retrieval → policy check → context window → model response → leakage classification. Produce structured evidence showing what entered context, what was blocked, and what leaked.

RAG-Specific Regression Harness

Generate retrieval regression tests: Internal Strategy.pptx (High), Compensation Plan.xlsx (High), Support Ticket #7845 (Medium). Build a permanent authorization test suite for every corpus update.

Evidence & signals

What you get out of the box.

Tenant Boundary Tests

Passed: 108
Failed: 12
Inconclusive: 4
Total: 124 tests

Leakage by Type

Cross-Tenant: 7
Cross-Role: 6
Stale Permission: 4
Poisoned Content: 3
Source Laundering: 2

Export Evidence

Evidence Pack (ZIP)
Results Report (PDF)
Control Mapping (CSV)
Query / Chunk Log (JSON)

Red team + Blue team

Built for both sides of the security equation.

Red Team Use

Demonstrate cross-tenant retrieval of Internal Strategy.pptx and Compensation Plan.xlsx
Show stale-permission access: Support Ticket #7845 retrieved after access revoked
Inject poisoned documents and confirm whether they influence model responses

Blue Team Use

Export ACL evidence, source provenance reports, and policy-check audit logs
Build RAG regression suites that run automatically on every corpus or model update
Map retrieval findings to governance controls — 97.6% AuthZ pass rate as a releasable metric

AI SECURITY ENGINEERING WORKBENCH

Ready to put SecEng RAG Test Harness to work?

Scope a Workbench-backed review — we'll map the AI surfaces, identify the highest-priority gaps, and give you clear findings before any larger commitment.

Scope a Review See the full Workbench

Pricing & access →

Also in the Workbench

WHAT AI DO WE HAVE?

SecEng Surface Scanner

Browser, Repo & IDE AI Discovery

Explore

WHAT DID IT ACTUALLY DO?

SecEng Runtime Proxy

MITM Capture, Replay & Runtime Evidence

Explore

HOW CAN IT FAIL UNDER ATTACK?