Free tools
Password Strength and Leak Checker
See whether a password appears in known breaches, and roughly how many guesses it would take to crack, with advice in NIST SP 800-63B terms. The estimate runs in your browser; the leak check asks Have I Been Pwned about 5 characters of the password's hash.
The password and the full hash never leave the browser. Nothing is stored, and nothing about what you type is counted or logged.
What the leak check sends, and to whom
When you press Check with the leak check on, this page hashes the password with SHA-1 in your browser and sends only the first 5 of the hash's 40 hex characters to Have I Been Pwned, at https://api.pwnedpasswords.com/range/ followed by those 5 characters. The browser first sends a CORS preflight: an OPTIONS request to the same /range/ address with no body, because of the Add-Padding header. Then it sends the GET, with the header Add-Padding: true.
The answer lists every hash ending in breach data that starts with those 5 characters, usually 800 to 1,000 of them, padded with dummy entries so its size gives nothing away. This page compares the other 35 characters of your hash with that list in the tab. The password and the full hash never leave the browser. No cookies and no referrer go with either request, and no other host is contacted. Like any cross-site request, both carry the browser's Origin header naming this site and its user agent. Have I Been Pwned and its network provider see your IP address, those 5 characters, the Origin and the user agent, as any web server sees a request.
Results
The leak check and the strength estimate appear here. They never show the password itself.
Reading the results
- Found in breaches
- The exact password is in the Pwned Passwords set: hashes of passwords from real breaches, with how often each one appears. Attackers load the same kind of data into their wordlists, so a password found here is tried early against every account, however unusual it looks. Stop using it. The count is rounded to a band, such as at least 1,000 times, because the exact figure is close to a fingerprint of the password.
- Not found
- Its hash was not in the set when you checked. That is useful, not conclusive: breaches that are not public yet, and passwords guessed from what someone knows about you, are not in any list.
- The strength estimate
- A count of the guesses an attacker who knows common patterns would need, priced by the cheapest way to build the password from words, runs, repeats, keyboard walks, years, dates and random characters. Very weak to very strong names that count, nothing more. A hit overrides the estimate: a password in breach data is very weak whatever its length. Words in a passphrase are priced as one of 7,776, the size of the EFF dice list, at most, and any other run of letters that reads like a word, such as a name, as one of 100,000, or at about one digit of guesses a letter when it is long. A phrase of three or more words is never priced above that many dice words.
- The three attack lines
- Online guessing, no lockout: About ten guesses a second against a login that never slows down or locks. NIST SP 800-63B has services allow no more than 100 failed attempts in a row on one account, which cuts an online attacker far below this. Offline, fast hash: About a hundred billion guesses a second: roughly one current high-end GPU against unsalted MD5 or NTLM, once a password database has been stolen. Offline, slow hash: About ten thousand guesses a second: roughly that same GPU against bcrypt at cost 10. Slow hashes exist to buy this difference.
- Length and character types
- Length is what NIST SP 800-63B measures. Character types change the price of the random parts a little, which is why one more random word does more than a capital and a symbol.
- Why results never show the password
- Findings name positions, such as characters 1 to 8, and kinds of pattern, never the characters themselves. Neither the hash prefix that was sent nor how many entries the answer held is shown: the real number of entries for a prefix is public, so it would point back to the prefix. Even so, the patterns say a lot about a weak password, so treat the results for a real one as you would the password.
What to do, in NIST SP 800-63B terms
- Length beats complexity. NIST SP 800-63B asks for at least 15 characters when the password is the only factor and at least 8 alongside MFA, and tells services not to impose rules such as one capital and one symbol.
- A passphrase of five or six words picked at random, by dice or a generator, is long and easy to type; four dice words fall to an offline fast-hash attack within days. Words that make a sentence you would say, or a line from a song or a book, are not random.
- A password manager makes a unique random password for every site and remembers it, so one breach stays one breach. NIST requires services to allow password managers and autofill, and says they should allow paste.
- Turn on MFA wherever it is offered, and prefer a passkey or a security key: they resist phishing, which a code by text message does not.
- Change a password when there is a sign it leaked, such as a hit here, not on a calendar. NIST tells services to stop forcing periodic changes and to refuse passwords found in breaches.
What this check cannot see
- Whether you use this password anywhere else. Reuse is how one breach opens other accounts, whatever the estimate says.
- The lists real attackers use. They run billions of leaked passwords and words with rules. This page knows a list of common passwords, keyboard walks, names and words, prices any other run of letters that reads like a word as one of 100,000 or more and a passphrase word as one of 7,776, so the estimate can still be too generous.
- Anything personal: names, pets, teams, places and dates that matter to you are guessed early by anyone who knows you or reads your profiles.
- How a site stores the password, whether it rate-limits logins, and whether MFA is on. Those decide which of the three attack lines applies.
- Breaches not yet in Pwned Passwords. Not found means the hash was not in that set when you checked, not that no attacker knows the password.
The strength figure is an estimate, from a model written for this page with a small built-in word list. Treat a high one as the absence of the obvious patterns, not as proof that the password will hold.
Where the leak data comes from
Leak counts come from Pwned Passwords, a free service of Have I Been Pwned, through its range API, which needs no key and was designed for this k-anonymity check. Its terms do not require attribution for Pwned Passwords and ask that the source be made clear, so it is credited here and in every result. Have I Been Pwned is a registered trade mark of Superlative Enterprises Pty Ltd. This site is not affiliated with it.
- Password spraying and MFA fatigue testing
- Password hash strength audit
- Hash Identifier, for which hashes are fast and which are slow