Password Strength Checker

Strength meters disagree because they measure different things

Most strength meters score composition: one uppercase, one digit, one symbol, eight characters minimum. That rule set comes from a 2003 NIST document, and it measures how annoying a password is to type rather than how hard it is to guess. P@ssw0rd1 satisfies every one of those rules and is in every cracking dictionary.

The number that actually matters is entropy — how many guesses an attacker needs. This tool reports two figures and explains the gap:

• Naive entropy — length × log₂(pool), where the pool is the set of character classes present. This is what a password looks like if an attacker knows nothing about how humans choose passwords.
• Adjusted entropy — the same figure reduced for the patterns a real attacker would exploit first: dictionary words, keyboard walks, sequences, repeats and dates.

Why the adjusted number is the one to trust

Attackers do not brute-force character by character. They run a dictionary with mutation rules, and those rules are extremely well tuned: password becomes P@ssw0rd, p4ssw0rd1!, password2024! and thousands of other variants, all tried in the first seconds. A twelve-character password built from one common word plus a suffix has a naive entropy that looks fine and an adjusted entropy that does not.

This is also why length beats complexity. Adding one random character multiplies the search space by the pool size; adding a symbol to a word an attacker already has does almost nothing, because the substitution rules cover it.

Password	Naive	Adjusted	Why they differ
password	38 bits	≈ 0	In every wordlist, first thing tried
P@ssw0rd1	59 bits	low	One word plus the standard leet substitution
qwerty123	47 bits	low	Keyboard walk plus a sequence
Tr0ub4dor&3	78 bits	moderate	Two words with substitutions — the famous xkcd example
correct horse battery staple	190 bits	high	Four random words; each is a full wordlist-sized choice

What the cracking time really means

The estimate assumes the attacker has the hash — a breached database — and can guess offline at full speed. That is the scenario that matters, because breaches happen and offline guessing has no rate limit. The three attacker models above differ by seven orders of magnitude, which is why "hard to guess online" and "hard to crack offline" are completely different claims.

The figure is an average of the search space, so it is the expected time to reach your password halfway through a systematic sweep. It is not a guarantee — an attacker can get lucky immediately, and a well-chosen wordlist can reach a weak password in seconds regardless of the theoretical size.

Frequently asked questions

Are long passphrases really better?

Yes, if the words are randomly chosen. Four words from a 256-word list is 32 bits; from a 7 000-word list it is about 50 bits and far easier to remember than an equal-strength random string. The randomness is the load-bearing part — a famous quote is a dictionary entry.

Why do sites still demand symbols and digits?

Habit and inherited policy. NIST SP 800-63B now recommends against composition rules and in favour of length, blocklisting known-bad passwords, and rate limiting. Many sites have not caught up.

Does a password manager change this?

Yes, decisively. If the password is generated randomly and never reused, length is free and every account can have an unrelated one. The main risk moves from guessing to the security of the manager itself.

Is my input sent anywhere?

No. Analysis runs in this page. Nothing is transmitted, logged or stored — not even to localStorage.