A Password Strength Checker With Entropy Math and Crack-Time Estimates

Strength = length × log₂(charset size) − penalties for common patterns. Crack time = 2^entropy / guesses_per_second. It's not zxcvbn, but it gets the basics right and shows users why their password is weak (bullet list of issues) instead of just a red bar with no explanation.

Most password strength meters are vibes-based: a bar that goes red/yellow/green based on opaque rules. Users learn to game the meter without actually making their passwords stronger. A meter that explains its reasoning — "you need a symbol", "this is in the top 500 common passwords", "you have 4 repeated characters" — is more educational.

🔗 Live demo: https://sen.ltd/portfolio/password-strength/
📦 GitHub: https://github.com/sen-ltd/password-strength

Features:

Strength bar with 5 levels
Entropy in bits
Crack-time estimate (seconds to centuries)
Specific improvement suggestions
500+ common password blocklist
Secure password generator (crypto.getRandomValues)
Passphrase mode (4-6 random words from 200+ wordlist)
Character class toggles, ambiguous character exclusion
Japanese / English UI
Zero dependencies, 67 tests

Entropy from character pool

export function calculateEntropy(password) {
  let poolSize = 0;
  if (/[a-z]/.test(password)) poolSize += 26;
  if (/[A-Z]/.test(password)) poolSize += 26;
  if (/\d/.test(password)) poolSize += 10;
  if (/[^a-zA-Z0-9]/.test(password)) poolSize += 32;
  if (poolSize === 0) return 0;
  return password.length * Math.log2(poolSize);
}

The formula assumes each character is chosen uniformly from the pool. It's an upper bound — real passwords are much less random, but it gives a consistent way to compare lengths × character classes.

Common pool sizes:

Lowercase only: 26 chars → 4.7 bits per char
+ Uppercase: 52 chars → 5.7 bits per char
+ Digits: 62 chars → 5.95 bits per char
+ Symbols: 94 chars → 6.55 bits per char

So a 12-character lowercase password has ~56 bits; adding uppercase, digits, and symbols brings it to ~79 bits. Adding two more characters to the all-classes version reaches ~92 bits, diminishing returns compared to just making it longer.

Crack time

export function crackTimeEstimate(entropy, guessesPerSec = 1e10) {
  const secs = Math.pow(2, entropy) / guessesPerSec;
  if (secs < 1) return 'instant';
  if (secs < 60) return `${Math.round(secs)} seconds`;
  if (secs < 3600) return `${Math.round(secs / 60)} minutes`;
  if (secs < 86400) return `${Math.round(secs / 3600)} hours`;
  if (secs < 31536000) return `${Math.round(secs / 86400)} days`;
  if (secs < 31536000 * 1000) return `${Math.round(secs / 31536000)} years`;
  return 'centuries';
}

10 billion guesses per second is a reasonable assumption for offline attacks against a properly-hashed password database (bcrypt/Argon2 slows this dramatically, but sha256 / md5 is in this range on modern GPUs).

At 10^10 guesses/sec:

40 bits → 110 seconds
50 bits → 31 hours
60 bits → 3.6 years
70 bits → 3700 years

The NIST threshold for "acceptable against offline attack" is about 64 bits. Most tools I've seen recommend 80+ for meaningful margin.

The common password blocklist

No amount of entropy math matters if your password is Password123! — it's in every cracking dictionary. A 500-entry blocklist catches the obvious cases:

export const COMMON_PASSWORDS = new Set([
  'password', '123456', 'qwerty', 'abc123', 'letmein', 'admin',
  'password1', 'Password1', 'Password123', 'P@ssw0rd',
  // ... 500 total
]);

export function isCommonPassword(password) {
  return COMMON_PASSWORDS.has(password.toLowerCase());
}

When the blocklist hits, the score is capped at 0 regardless of entropy. A long "common" password (correcthorsebatterystaple — actually famously used in XKCD) scores well on entropy but low on novelty.

Sequential and repeated patterns

export function hasSequential(str) {
  const s = str.toLowerCase();
  for (let i = 0; i < s.length - 2; i++) {
    const c0 = s.charCodeAt(i);
    const c1 = s.charCodeAt(i + 1);
    const c2 = s.charCodeAt(i + 2);
    if (c1 === c0 + 1 && c2 === c1 + 1) return true; // abc, 123
    if (c1 === c0 - 1 && c2 === c1 - 1) return true; // zyx, 321
  }
  return false;
}

export function hasRepeated(str) {
  return /(.)\1\1/.test(str); // aaa, 111
}

These patterns reduce effective entropy significantly — qwertyuiop has theoretical ~47 bits but dictionary attacks crack it in milliseconds. Detecting and flagging them feeds into the bullet-list feedback.

Crypto-secure generation

export function randomInt(max) {
  const buf = new Uint32Array(1);
  crypto.getRandomValues(buf);
  return buf[0] % max;
}

Math.random() is NOT suitable for password generation — it's seeded from the current time and is predictable. crypto.getRandomValues draws from the OS entropy pool. In Node 18+ and all modern browsers, it's available on globalThis.crypto.

Passphrases

The XKCD approach: 4 random words = ~52 bits if your wordlist has ~8000 entries:

export function generatePassphrase(wordCount, separator, wordlist) {
  const words = [];
  for (let i = 0; i < wordCount; i++) {
    words.push(wordlist[randomInt(wordlist.length)]);
  }
  return words.join(separator);
}

My wordlist has 200+ entries. Smaller than EFF's 7776-word list (~12.9 bits/word) but still produces memorable passwords at ~7.6 bits/word. At 4 words that's ~31 bits — too weak. At 6 words, ~46 bits — borderline. Users should prefer the character-based generator for anything sensitive, but passphrases have their place for typed credentials.