Password Generator

How to use Password Generator

1. Set the length and choose which character sets to include, or switch to passphrase mode.
2. Read the entropy in bits and the estimated crack time.
3. Copy the password. It is generated locally and never transmitted.

How long should a password be?

Length beats complexity. Each additional character multiplies the search space; adding a symbol to a short password barely moves it.

The practical recommendation for 2026: 16 characters minimum for anything you type rarely and store in a manager, or a 6-word passphrase for anything you must type by hand — a device login, a password manager master password, a disk encryption key.

What is entropy and why does it matter?

Entropy measures how many guesses an attacker needs, expressed in bits. Each bit doubles the search space, so 60 bits is not twice as strong as 30 bits — it is a billion times stronger.

It is calculated as log₂(pool size) × length. A 12-character password from the 94 printable ASCII characters gives log₂(94) × 12 ≈ 78.7 bits.

The critical condition is that the password must actually be random. "P@ssw0rd123!" has 12 characters from a 94-character pool, which naively computes as 78 bits — but it is in every cracking dictionary and falls in under a second. Entropy only counts if the selection was genuinely random, which is exactly what a generator provides and a human choosing a memorable password does not.

Are passphrases better than random passwords?

For anything you have to type from memory, yes. For anything stored in a password manager, it makes no difference.

A passphrase built from randomly selected dictionary words — the Diceware method — gets its entropy from the word list size, not the character count. A 7,776-word list gives 12.9 bits per word, so six words is 77.5 bits: comparable to a 12-character random password, and vastly easier to remember and type on a phone or a TV remote.

The requirement is that the words be chosen randomly by the generator. Picking six words yourself produces something close to a sentence, and human-chosen word sequences follow predictable patterns that cut the effective entropy by more than half.

The XKCD "correct horse battery staple" example is four words at 44 bits, which was reasonable in 2011 and is marginal now. Use six.

Why does Math.random() matter?

Because it is not random enough for anything security-relevant, and a surprising number of online password generators use it.

Math.random() is a pseudorandom number generator optimized for speed. In V8 it uses xorshift128+, seeded from a limited entropy source, and its internal state can be recovered from a modest number of observed outputs. Given a few generated values, an attacker can predict every subsequent one.

crypto.getRandomValues() draws from the operating system’s cryptographically secure entropy pool — /dev/urandom on Unix, BCryptGenRandom on Windows — which is designed to be unpredictable even to an adversary who has seen prior output.

This tool uses the latter, with rejection sampling to avoid the modulo bias that would otherwise make some characters marginally more likely than others. It is a small detail, and it is the difference between a uniform distribution and a slightly skewed one.

Frequently asked questions