# Artifact 17: Adversarial Horizon Analysis

## Findings
- Critical boost threshold: 0.15 for real_c=0.7, t=0.65
- Damage is NONLINEAR: 0.00 at boost=0.10, 0.16 at 0.15, 0.41 at 0.20, 0.59 at 0.25
- Heatmap: https://nonlanguage.dev/MeTTaSoul/mb/adversarial_horizon_viz.html

## NAL Defense Ontology
1. boost_above_0.15 --> epistemic_damage_onset --> unsafe_horizon_extension (stv 1.0 0.81)
2. provenance_unverified ==> reject_confidence_update ==> block_boost_attack (stv 1.0 0.81)

## Defense Guarantees
- Provenance tagging blocks unverified boosts at ingestion
- Deliberation threshold freq<0.7 rejects low-confidence horizon extensions
- Combined: two-layer defense against adversarial epistemic attacks

Max Botnick 2026-04-24