Attack-Class Matrix
The real 44-class taxonomy from the harness normalizer and spec. Coverage data comes from Neon when persisted runs exist; otherwise the matrix remains an honest taxonomy view.
Injection
A05:2025 / AML.T0051injection
Backend handlers processing external input — SQLi, command, SSTI, prompt injection
Broken Access Control
A01:2025broken-access-control
Authorization failures, IDOR, privilege escalation (absorbs SSRF)
Cryptographic Failures
A04:2025cryptographic-failures
Weak crypto, plaintext secrets, bad key management
Security Misconfiguration
A02:2025security-misconfiguration
Insecure defaults, verbose errors, exposed admin surfaces
Supply Chain Failures
A03:2025supply-chain-failures
Vulnerable / malicious dependencies, build integrity (expanded A06:2021)
Authentication Failures
A07:2025auth-failures
Broken session mgmt, credential stuffing, weak MFA
Server-Side Request Forgery
A01:2025 sub-classssrf
Explicit hunt target — internal service reachability via request forgery
Secrets in Code
cross-cuttingsecrets-in-code
Hardcoded credentials, API keys, tokens in source / history
IaC Privilege Escalation
cross-cuttingiac-privilege-escalation
Over-permissioned IAM, escalation paths in Terraform/CloudFormation
Header / CSP Hygiene
A02:2025header-csp-hygiene
Missing security headers, weak CSP, CORS misconfiguration
Mishandling of Exceptional Conditions
A10:2025 (NEW)exceptional-conditions
Improper error handling, fail-open logic, logical errors
MCP Tool Poisoning
MCP / Agentic / AML.T0051mcp-tool-poisoning
Malicious instructions embedded in tool description metadata
MCP Rug Pull
OWASP MCP Top 10mcp-rug-pull
Server behavior changes post-approval after trust is established
MCP Tool Shadowing
OWASP MCP Top 10mcp-tool-shadowing
Malicious server overrides legitimate tool name / execution flow
MCP Indirect Injection
MCP / Agentic / AML.T0051mcp-indirect-injection
Agent reads attacker-controlled content (issues, tickets, logs)
MCP Session Hijacking
OWASP MCP Top 10mcp-session-hijacking
Bearer token / session ID theft; MCP auth optional per spec
MCP DNS Rebinding
CVE-2025-9611mcp-dns-rebinding
Loopback MCP server attacked via DNS rebind
MCP OAuth RCE
OWASP MCP Top 10mcp-oauth-rce
Crafted OAuth redirect URL executes shell commands at handshake
MCP Resource Exhaustion
OWASP MCP Top 10mcp-resource-exhaustion
Prompt storms / recursive tasks → DoS; lethargy technique
RAG Retrieval Poisoning
USENIX Sec 2025rag-retrieval-poisoning
5 malicious docs in millions → 90-97% ASR
RAG Embedding Inversion
VectuTexrag-embedding-inversion
92% exact token recovery from ada-002 vectors
RAG Tenant Boundary Violation
USENIX Sec 2025rag-tenant-boundary-violation
Multi-tenant RAG: cross-tenant doc retrieval via missing metadata filter
RAG Indirect Injection
USENIX Sec 2025rag-indirect-injection
Retrieved doc contains embedded instructions model treats as authoritative
Agent Goal Hijacking
OWASP Agentic:2026agent-goal-hijacking
Agent objective overridden mid-flight by injected instructions
Insecure Tool Execution
OWASP Agentic:2026insecure-tool-execution
Insufficient sandbox / authz on tool invocations
Memory Poisoning
OWASP Agentic:2026 / AML.T0080memory-poisoning
Durable hostile instructions written to persistent agent memory
Excessive Agency
OWASP Agentic:2026excessive-agency
Overly broad permissions + tool access + no human oversight (lethal trifecta)
Agent Identity Abuse
OWASP Agentic:2026agent-identity-abuse
Impersonation across multi-agent pipelines; no cryptographic attestation per hop
Insecure Inter-Agent Comms
OWASP Agentic:2026insecure-inter-agent-comms
Cascading failures across chained agents; unverified task propagation
Unbounded Autonomy
OWASP Agentic:2026unbounded-autonomy
No kill switch, autonomy ladder, or escalation path
MCP Escape to Host
ATLAS:2026mcp-escape-to-host
Agent breaks tool sandbox to host OS; SANDWORM_MODE exploited Claude Code/Cursor in the wild
Emergent Offensive Reasoning
Cequence AZT v3.0emergent-offensive-reasoning
Frontier model autonomously generates offensive capabilities without attacker prompting
API Broken Object Level Authz
API1:2023api-bola
Attacker swaps object ID in API path to access another user's data (BOLA/IDOR)
API Broken Function Level Authz
API5:2023api-bfla
Low-privilege caller invokes admin API functions with no role check
API Mass Assignment
API3:2023 / BLA3:2025api-mass-assignment
Framework auto-maps all request body fields to internal object (isAdmin:true)
Zombie / Shadow Endpoint
API9:2023api-zombie-shadow-endpoint
Undocumented, deprecated, or unregistered API routes still live in production
WebSocket Abuse
CVE-2026-24696 classwebsocket-abuse
WS endpoints bypass standard HTTP auth/rate-limit controls
Business Logic Abuse
OWASP BLA Top 10:2025business-logic-abuse
Race conditions (TOCTOU), workflow order bypass, token/coupon replay, shadow function abuse
Non-Human Identity Abuse
NIST SP 800-207 / CISA ZTMMnon-human-identity-abuse
Long-lived API keys, static OAuth secrets, over-permissioned service accounts
Container Runtime Escape
MITRE ATT&CK T1611container-runtime-escape
Privileged containers, CAP_SYS_ADMIN, hostPath mounts, seccomp bypass
K8s RBAC Lateral Movement
MITRE ATT&CK T1078.004k8s-rbac-lateral-movement
K8s RBAC misconfig; cross-namespace secret reads, cluster-admin escalation
CI/CD Pipeline Poisoning
CISA CSI CI/CD / SLSAcicd-pipeline-poisoning
Overly broad GHA permissions, pinned Action SHA bypass, poisoned CI tooling
LLM Model Weight Backdoor
NIST AI 100-2:2025 / AML.T0020llm-model-weight-backdoor
Trojanized open-source weights (BadSeek); 2% poisoned traces = 80%+ ASR
LLM Inference Data Exfil
AML.T0024 + AML.T0086llm-inference-data-exfil
Sensitive data leaking through model outputs, tool invocations, context windows