- Overview
- Attack Efficiency Analysis
- Invisible Exposure Matrix
- Observation-to-Disclosure Model
- Synthetic Scenario Lab
- Evidence & Provenance
- Research Methodology
Invisible Exposure Matrix
Version 0.1 records monitoring coverage. It does not estimate hidden attack counts.
UNKNOWN remains UNKNOWN. A quiet sensor is not evidence of safety, and an unverified exposure is not an active compromise.
| Surface | Exposure | Coverage | Evidence | Provenance |
|---|---|---|---|---|
AEIE-X-01 Public Web Applications Internet-reachable web applications, including member, reservation, and administrative surfaces. | UNKNOWN | UNKNOWN | VERIFY HOLDinsufficient evidence | UNKNOWN Last observation: UNKNOWN |
AEIE-X-02 Backend APIs API endpoints that serve web or mobile clients, including endpoints not linked from a public UI. | POTENTIAL_EXPOSURE | PARTIAL | INFERENCE | PUBLIC_SOURCE Last observation: 2026-10-15 |
AEIE-X-03 API Authorization Boundaries Object- and function-level authorization that must hold even when a request looks well-formed. | POTENTIAL_EXPOSURE | UNKNOWN | INFERENCE | PUBLIC_SOURCE Last observation: 2026-10-15 |
AEIE-X-04 Machine Credentials Access keys, API keys, root accounts, and long-lived tokens. | POTENTIAL_EXPOSURE | UNKNOWN | STATED | PUBLIC_SOURCE Last observation: 2026-10-15 |
AEIE-X-05 Cloud Service Permissions Cloud IAM and service quotas that turn a credential into billable or data-moving authority. | UNKNOWN | UNKNOWN | VERIFY HOLDinsufficient evidence | UNKNOWN Last observation: UNKNOWN |
AEIE-X-06 AI Agent Tool Permissions Tools, connectors, and credentials available to agents once they can act. | UNKNOWN | UNKNOWN | EXCLUDE | PUBLIC_SOURCE Last observation: 2026-10-15 |
AEIE-X-07 Shadow AI Integrations AI services and agents connected outside a supervised inventory. | UNKNOWN | UNKNOWN | VERIFY HOLDinsufficient evidence | RESEARCH_OBSERVATION Last observation: UNKNOWN |
AEIE-X-08 Third-Party Service Connections Processors, re-processors, and connected SaaS that inherit data or credentials. | UNKNOWN | UNKNOWN | VERIFY HOLDinsufficient evidence | UNKNOWN Last observation: UNKNOWN |
AEIE-X-09 Logging and Telemetry Whether attempts, API access, and authentication events are retained long enough to investigate. | UNKNOWN | UNKNOWN | VERIFY HOLDinsufficient evidence | UNKNOWN Last observation: UNKNOWN |
AEIE-X-10 Disclosure Visibility The subset of confirmed or suspected incidents that become public reports. | UNKNOWN | PARTIAL | STATED | PUBLIC_SOURCE Last observation: 2026-10-15 |
Exposure coverage, not hidden-incident estimates
Unit: domain / status label
Time window: module v0.1; LAC access 2026-10-21
Measurement scope: Research matrix. No live asset inventory.
Data origin: UNKNOWN coverage unless a public source is cited in-row
EMPIRICAL / SOURCE-BACKED
AEIE-X-01 · Public Web Applications
LAC notes that internet-facing assets beyond marketing sites are in scope for current Japanese disclosures. That is a public-source observation, not a verified exposure of a named system here.
Limitation: This Observatory does not scan third-party web applications. No exposure found is not a finding.
AEIE-X-02 · Backend APIs
NICT, as cited by LAC, describes App-to-Backend analysis as a confirmed attack path in some Japanese leakage cases. AEIE does not enumerate those APIs.
Limitation: Potential exposure is a research category for surfaces that can exist without frontend visibility. It is not a confirmed compromise.
AEIE-X-03 · API Authorization Boundaries
LAC notes that conventional WAF signatures may not stop authorized-looking requests that exceed object-level permission.
Limitation: IDOR/BOLA cannot be inferred from frontend permission labels. Coverage of authorization checks is UNKNOWN here.
AEIE-X-04 · Machine Credentials
Stolen credentials used for Vertex AI / Bedrock consumption appear in LAC's handled-case narrative as an extension of credential compromise, not as a new attack class.
Limitation: LAC reports credential misuse in handled cases. That is not a census of key hygiene.
AEIE-X-05 · Cloud Service Permissions
Authority Surface already treats capability and permission as distinct. AEIE records the monitoring gap only.
Limitation: No cloud tenant is in AEIE's observation scope.
AEIE-X-06 · AI Agent Tool Permissions
Tool permission remains an Authority Surface question. It is not filled by overseas campaign reports.
Limitation: LAC states there is no public evidence of AI-agent involvement in the Japanese series. Unsupported attribution is excluded.
AEIE-X-07 · Shadow AI Integrations
Observability of unmanaged AI services is a coverage problem. Unknown coverage stays UNKNOWN.
Limitation: This matrix does not discover shadow agents. See Shadow AI Agent Supervision.
AEIE-X-08 · Third-Party Service Connections
LAC recommends periodic, not only contractual, review. That is operational advice, not an AEIE finding.
Limitation: Supply-chain surfaces are not inventoried by AEIE v0.1.
AEIE-X-09 · Logging and Telemetry
Declining sensor coverage can lower observed signal volume with unchanged underlying activity. Scenario C models that distortion synthetically.
Limitation: If logs are absent, attack activity is not shown to be absent. The correct state is UNKNOWN.
AEIE-X-10 · Disclosure Visibility
Mean disclosure latency in LAC's dated subset fell from 10.9 days (May 2026) to 2.9 days (September 2026).
Limitation: LAC's disclosure series is a public-reporting sample. Faster reporting can raise counts without raising attacks.