Module

ASEMA

Automated Security Assessment. Multi-cycle analysis with triple modular redundancy and deterministic scoring. Built for evaluating messaging platforms in classified-compatible environments.

7
Security Doors
63
Total Controls
3
Independent Models
3
Platforms

Architecture

ASEMA evaluates messaging platform security through 7 "doors" — each representing a security domain (authentication, encryption, data sovereignty, access control, compliance, audit, and operational security). Each door contains multiple controls, totaling 63 across the framework.

Triple Modular Redundancy: three independent reasoning engines evaluate each control separately. Scores are reconciled through deterministic consensus. If two of three agree, the score holds. If all three disagree, the control is flagged for human review.

Scoring

Deterministic scoring: WEIGHTED_COVERAGE_%. Each control receives a binary or graded assessment. Controls are weighted by criticality within their door. Doors are weighted by overall security importance. Final score is fully decomposable — you can trace any percentage point back to a specific control assessment.

Anti-Hallucination

4-layer constraint system prevents model confabulation:

  • Schema enforcement — responses must match predefined JSON schemas. Non-conforming output is rejected.
  • Citation requirement — every assessment must reference specific platform documentation or observable behavior.
  • Cross-model validation — TMR catches single-model hallucinations through majority consensus.
  • Deterministic aggregation — final scores computed arithmetically from validated sub-scores, not generated by models.

Platforms

Microsoft Teams
Enterprise collaboration platform. Evaluated for DoD and federal compliance.
Slack
Team messaging platform. Assessed against enterprise security requirements.
Signal
End-to-end encrypted messaging. Evaluated for operational security contexts.

Deployment

Fully offline. Classified-compatible. All models run locally. No platform data leaves the assessment environment.