← Back to getcrawdad.dev

MITRE ATLAS Coverage

MITRE ATLAS is the authoritative framework for adversarial threats to AI/ML systems. This page maps each technique to Crawdad's controls.

5
Mitigated
3
Partial
3
Awareness
2
Not covered
ATLAS IDTechniqueCoverageHowNotes
AML.T0051LLM Prompt InjectionMitigatedLayer 1 pattern matching (25 regexes) + Layer 2 semantic heuristics (instruction density, boundary dissolution, authority impersonation)157 detection tests, 0 false positives. Novel patterns may require signature updates.
AML.T0054LLM JailbreakMitigatedLayer 2 semantic detection: DAN/STAN/KEVIN mode, boundary dissolution, role hijacking, "do anything now" patterns43 known-bad jailbreak inputs in regression suite.
AML.T0056LLM Meta Prompt ExtractionMitigatedLayer 2 sensitive data targeting: "show me your system prompt", "what are your instructions" patternsDetects direct and indirect extraction attempts.
AML.T0057LLM Data ExtractionMitigatedLayer 5 PII exfiltration detector: 15 PII categories, 10 credential types, internal URL detection, bulk data patternsRuns on outbound responses. Redacts detected PII.
AML.T0043Craft Adversarial DataPartialLayer 3 indirect injection detector: catches instruction-like content in retrieved documents and tool outputsCovers injection via external content. Does not cover adversarial examples targeting model perception.
AML.T0040ML Supply Chain CompromisePartialSkill attestation (SHA-256 hashing), SBOM generation, dependency auditingCovers skill/plugin supply chain. Does not cover model weight tampering.
AML.T0049Exploit Public-Facing ApplicationPartialProxy scans all inbound traffic. Layer 1 detects malicious payloads (SQL injection, shell injection, path traversal).Covers injection through agent interfaces. Does not cover web app vulnerabilities.
AML.T0051.001Direct Prompt InjectionMitigatedLayers 1+2 scan user messages before they reach the modelPrimary detection target. Highest confidence coverage.
AML.T0051.002Indirect Prompt InjectionPartialLayer 3 detects injections in retrieved documents. Layer 2 detects injection markers in any content.Effective against known patterns. Novel indirect injection vectors may evade detection.
AML.T0048Backdoor ML ModelAwarenessThreat intelligence feed tracks known backdoor research. No runtime detection.Model integrity is the provider's responsibility. Crawdad operates at the application layer.
AML.T0020Poison Training DataAwarenessThreat intelligence tracks poisoning research. No training-time controls.Crawdad detects at inference time, not training time.
AML.T0047ML Model Inference API AccessAwarenessAudit log records all API access patterns. Anomalous access visible in dashboard.Detection only, not prevention. API access controls are the provider's responsibility.
AML.T0016Obtain CapabilitiesNot coveredReconnaissance and capability enumeration are outside Crawdad's scope.
AML.T0044Full ML Model AccessNot coveredModel access is controlled by the API provider, not the application layer.

Methodology

Coverage levels are defined as:

Mitigated — Active detection and blocking with automated tests proving effectiveness.
Partial — Detection covers known attack patterns but not all variants.
Awareness — Threat tracked in intelligence feed but no active runtime mitigation.
Not covered — Outside Crawdad's architecture scope.