Controls
Presets
Parameters
Quality
Camera
Model checks
The science
Why WPA2 can be attacked offline. In WPA2-Personal the pairwise master key is PMK = PBKDF2-HMAC-SHA1(passphrase, SSID, 4096, 256 bit). Anyone who records one 4-way handshake can test a candidate passphrase on their own hardware: derive the PMK, recompute the handshake check, compare. No contact with the access point is needed, so the only limit is hardware speed G (guesses per second). Each guess costs 4096 × 2 HMAC-SHA1 calls, which is why PBKDF2 slows attackers down but does not stop a GPU farm.
Why WPA3 changes the game. WPA3-Personal replaces the handshake with SAE (Dragonfly), a password-authenticated key exchange. A recorded exchange does not let an observer test guesses; every guess needs a live protocol run with the access point, which can rate-limit and lock out. With g attempts per second, k failures allowed before a lockout of tl seconds, the long-run rate is r = k / (k/g + tl).
People are not random. An attacker tries likely passphrases first. Modelling choices as a Zipf list, P(rank r) ∝ r-s over N candidates, the chance of success after M guesses is H(M,s)/H(N,s) with H(M,s) = ∑r≤M r-s (the sum is evaluated by Euler-Maclaurin for large M). With strong skew a short, predictable passphrase falls to a few thousand guesses, however big the list is; a truly random passphrase of the same list size needs about N/2 guesses.
Simplifications. One hash rate for the whole attack; the list order is known to the attacker; no dictionary-plus-rules mutations; WPA3 side-channel flaws (Dragonblood) and transition-mode downgrade to WPA2 are not modelled (they weaken real deployments, so treat WPA3 results as the ideal protocol); an offline attacker can also capture a handshake only when a client joins. Defensive explanation only, no tooling. Checks: the real IEEE 802.11i PBKDF2 test vector computed by an in-page SHA-1, Euler-Maclaurin sums against direct summation, the quantile sampler against the formula, and an event simulation of the lockout.