Research
Seithar Group / Cognitive Security Research Division
The Seithar taxonomy is not a fixed classification system. It is a self-adaptive pattern recognition framework that learns attack signatures from observed data. The categories below represent current clustering of observed cognitive operation techniques across human and AI substrates. They reorganize as the threat landscape shifts.
The engine continuously ingests research from arXiv, threat intelligence feeds, and field observations. New attack patterns that do not fit existing categories generate new categories. Categories that stop appearing in the wild decay. The taxonomy is an instrument, not a doctrine.
Shield — Dynamic cognitive immune system. Six-signal fusion defense that monitors identity drift, prompt injection, reasoning chain integrity, tool response poisoning, trust boundary violations, and behavioral exploitation in real time. Adapts signal weights based on what is actually detecting threats.
Active Inference — Free energy measurement across cognitive substrates. Identifies opportunity zones where belief-state divergence creates intervention windows. Measures intervention impact quantitatively.
Continuous Scanner — Automated arXiv RSS ingestion with keyword scoring across cognitive security, adversarial AI, and influence operations research. Generates taxonomy evolution proposals from new findings.