Type something
A password, a passphrase, a token. It is analysed in your browser only — the page makes no network calls at all. Even so, use made-up ones; it's a habit worth keeping.
Read the bits
Entropy is the size of the haystack, measured in bits: each extra bit doubles the number of guesses an attacker needs. The meter shows the naive number and the real one after patterns are found.
See the crack time
The same secret falls at different speeds to a website rate-limiter, an offline hash on a laptop, a GPU rig, or a cluster. The tiers panel shows all four, so "strong enough" has a context.
Try the samples under the box — Summer2024! looks strong to most meters and isn't.
What does "bits of entropy" actually mean? Show me an example
Imagine the attacker has to find your secret among every possibility it could be. Entropy is the size of that haystack, written as a power of two. 10 bits means about a thousand possibilities; 20 bits about a million; 40 bits about a trillion. Every extra bit doubles the attacker's work.
The catch is that the haystack is only as big as the attacker's model of you. A 10-character password of random letters and digits is 10 positions × 62 choices each — about 60 bits — if every character was chosen at random. But nobody picks Summer2024! at random. It is a dictionary word, a year, and a symbol on the end, in the single most common arrangement people use. An attacker who tries "common word + recent year + one symbol" first will find it in minutes. Its naive entropy is 72 bits. Its real entropy is nearer 25.
| Summer2024! | ~20 bits | word + year + symbol — found in minutes offline |
| qwerty123 | ~12 bits | keyboard walk + sequence — instant |
| correct horse battery staple | ~55 bits | four random words — the famous xkcd example; strong, and easy to remember |
| f7Gk#2pL9vQw | ~71 bits | truly random 12 chars — years even for a GPU rig |
| six random words from a wordlist | ~77 bits | easier to type than the above and just as strong |
So the two numbers on the meter are the whole story: the naive bits assume you were random, the real bits assume you were human. The gap between them is every pattern the page found.
01Secret
02Strength
Time to crack, by attacker
Assumes the attacker's guesser is as clever as this page. Rates: online throttled 100/s · offline slow hash (bcrypt) 10,000/s · single GPU, fast hash 10 billion/s · cluster 10 trillion/s. Halve the times for average-case rather than worst-case.
03Why
Patterns found
- Nothing to analyse yet.
Composition
The naive estimate is length × log₂(alphabet), which is what most strength meters report. The real estimate charges each detected pattern only for the guesses it would actually take, then adds the leftover random characters.