Password Generator

How the Randomness Actually Works

This generator draws its randomness from window.crypto.getRandomValues() — the browser's cryptographically secure pseudorandom number generator (CSPRNG) — rather than Math.random(). That distinction matters: Math.random() is specified for statistical randomness, not unpredictability, and its output has been reverse-engineered in some browser engines to predict future values from past ones. crypto.getRandomValues() is backed by the operating system's secure entropy source and is the right primitive for anything security-sensitive, including passwords.

One small technical nuance worth knowing: each character is picked with randomValue % charSet.length. When the character set's length doesn't evenly divide 2³², this introduces a theoretical modulo bias — some characters become infinitesimally more likely than others compared to true uniform selection (the mathematically pure approach is rejection sampling, discarding out-of-range values instead of wrapping them with modulo). In practice, for character sets in the tens of characters against a 32-bit random source, this bias is many orders of magnitude smaller than other real-world risks to a password's security — it's a fact worth knowing if you're evaluating the algorithm, not a reason to distrust the output.

What Actually Determines Brute-Force Resistance

A password's resistance to brute-force guessing is measured in bits of entropy, calculated as length × log₂(character set size) — not by how "random-looking" it is to a human eye. With all four character classes enabled here (26 lowercase + 26 uppercase + 10 digits + 26 symbols = 88 possible characters per position), a 16-character password carries roughly 103 bits of entropy (16 × log₂88 ≈ 16 × 6.46). Dropping to a 12-character password with the same character set drops you to about 77 bits — still strong, but the difference compounds fast because entropy scales with length, not just character variety. Turning off a character class (say, disabling symbols) shrinks the set to 62 characters and costs roughly 1 bit of entropy per position compared to the full 88-character set — a small-looking change that adds up across a long password.

Why "Exclude Ambiguous Characters" Exists — and What It Actually Removes

This option strips exactly seven characters from the pool: i l 1 L o 0 O — the lowercase/uppercase/digit trios that render nearly identically in many fonts (lowercase L vs. digit 1, uppercase O vs. digit 0). It exists for one specific, human problem: when a person has to type a generated password by hand — reading it off a screen, a printed card, or dictating it over a phone — these characters cause real transcription errors. It does not improve resistance to automated brute-force guessing at all (a computer doesn't confuse "l" and "1"); it purely trades a small amount of entropy for fewer human typos. If the password will only ever be pasted from a password manager and never hand-typed, there's no security reason to enable it.

The strength meter shown after generation is a simple heuristic — it awards points for length crossing 8/12/16 thresholds and for each character class present, capped at 100. It is not a true entropy calculation and can rate two passwords of very different actual brute-force resistance as the same "Very Strong" bucket. Use the entropy formula above if you need an actual comparable security number rather than the bar's qualitative label.

Frequently Asked Questions