Password strength checker

How this checker estimates strength

Many password meters count character types and hand out points for a capital letter or a symbol. That rewards P@ssw0rd1!, one of the first passwords any attacker tries. This checker uses zxcvbn, the estimator Dropbox released as open source in 2012 and described in a peer-reviewed paper at USENIX Security 2016, in its maintained TypeScript version, zxcvbn-ts. Instead of counting symbols, it asks how many guesses a cracking tool would need.

It splits the password into the pieces an attacker would guess separately: entries from lists of common passwords, English words, first names and surnames, dates and years, keyboard runs such as qwerty or zaq1, sequences such as abcd or 4321, repeats, and look-alike swaps such as @ for a. Each piece gets a guess count, and the cheapest way to cover the whole password becomes the estimate. Anything that matches no pattern counts as random characters, the most expensive kind to guess.

The four attack speeds

• Login page with limits (100 guesses an hour). Someone trying passwords at a site that slows down or locks out failed logins.
• Login page without limits (10 a second). A script hammering a site that never blocks it.
• Stolen database, slow hash (10,000 a second). The site was breached, but stored passwords with a deliberately slow hash such as bcrypt, scrypt or Argon2.
• Stolen database, fast hash (10 billion a second). The site used a fast hash such as MD5 or SHA-1, and the attacker has a rack of graphics cards.

You can't control how a site stores your password, so aim for one that survives the fast-hash row for years, not minutes. Here is how some typical passwords fare, rated by this checker:

The examples on this page are public now, so don't use them. Notice how little the capital letters, digits and symbols in P@ssw0rd and Summer2024! help, and how much random words and length do.

One caution about random passwords: zxcvbn can't tell a truly random string from one that merely looks random, so it assumes only about 10 possibilities for each character it can't explain. A generator actually picks from 70 to 90 characters, so the times above for random passwords are deliberately pessimistic. For a password you generated, the password generator shows its exact strength.

Why length beats complexity

NIST's digital identity guidelines (Special Publication 800-63B, revised in 2025) set the rules many organizations follow. They require passwords used on their own to be at least 15 characters long (8 when the password is just one part of multi-factor sign-in), ask sites to accept at least 64 characters, and tell them not to impose composition rules such as "one uppercase letter, one number and one symbol". Sites must instead check new passwords against a blocklist of common, expected and breached ones, and shouldn't force periodic changes unless a password may have been compromised.

The reason is human behavior. Faced with complexity rules, most people capitalize the first letter and put a number and an exclamation mark at the end, which cracking tools try early. Every extra character of genuine randomness multiplies the work, so a long passphrase of random words such as lagoon skeptic blissful canteen beats a short, cluttered password, and it's far easier to type.

How to fix a weak password

• Let a machine choose it. Use the password generator for random characters, or the passphrase generator if you have to type it.
• Make it longer, with unpredictable words rather than a year or an exclamation mark.
• Use it on one site only, and keep all your passwords in a password manager.
• Check it hasn't leaked. A strong-looking password that appears in breach data is unsafe; the password leak check tells you without revealing it.
• Turn on two-step verification for email, banking and anything important, so a password alone isn't enough.