Password entropy. A secret picked uniformly at random from N possibilities carries H = log2 N bits. L characters from an alphabet of size A give H = L log2 A (8 lowercase letters: 37.6 bits); k words from a 7776-word list give 12.92 k bits. On average an attacker must try (N + 1)/2 ≈ 2H−1 guesses, so every extra bit doubles the work.
Work factor. A slow, salted password hash repeats its core 2c times, dividing the guess rate by 2c. The page uses an illustrative 1010 guesses per second for a fast hash on dedicated hardware; real rates depend on the algorithm and hardware and are only order-of-magnitude figures here. Online services add a different defence: lockout after a limited number of tries, so a random secret falls with probability tries / 2H.
Defence in depth. If layers fail independently, an intruder must pass all of them, so the chance of reaching the core is the product of the per-layer bypass probabilities (the page checks this against a seeded Monte Carlo run). The five rings map to the CIA triad: confidentiality (password, MFA), integrity (segmentation, patching) and availability (monitoring/response, backups). Backups reduce the damage, not the likelihood. Risk = likelihood × impact, here the expected loss per year = attempts × breach probability × loss.
Simplifications. The layer probabilities are teaching values, not measurements; real layers are correlated (one flaw can defeat two defences), which makes the product too optimistic. Passwords are assumed truly random, but people choose far weaker ones, and reuse and phishing matter more than brute force. This is a concept model of probabilities on an imaginary vault: it contains no exploit, scanner or attack tool and does not touch any real system.
Running numerical checks...