Fig. 01 — Model structure (5 stocks, 3 feedback loops)
τ_d + τ_m + B1 — response loop − R1 + activation + B3 — adaptation loop − φ tampering + B2 — recovery − A attacker foothold P perceived compromise R response capacity E containment effectiveness B ballast deviation

Amber marks the reinforcing loop (R1) — attacker propagation left unchecked. Cyan marks the response loop (B1) — perception, delayed by τ_d, mobilises response that contains the attacker. Purple marks the adaptation loop (B3) — sustained response gradually raises containment effectiveness. Ballast deviation (B, red) is the physical consequence: it has no feedback of its own, which is exactly why it can only be managed indirectly, through A and R.

Fraction of system state — foothold, perceived compromise, response

Attacker foothold A Perceived compromise P Response capacity R

Ballast water-level deviation vs. critical threshold

Ballast deviation B B_crit (stability loss)