How does Web Crypto generate secure passwords in your browser?
Every password and passphrase in this workspace is drawn from your browser’s cryptographically secure pseudo-random number generator (crypto.getRandomValues) rather than Math.random(). For true in-browser password generator privacy, Web Crypto pulls entropy directly from the operating system kernel without any server round-trip.
As documented in the MDN Web Crypto API Crypto.getRandomValues() specification, typed arrays filled by the browser provide uniform random integers suitable for cryptographic keys and session tokens. If you also need unique identifiers for API keys or database records alongside login credentials, you can use uuid generator in your browser with the same local randomness.
Rejection sampling vs modulo bias
Taking a random 32-bit integer modulo an alphabet length introduces slight probability bias whenever the alphabet size does not divide 2^32 evenly. The underlying randomInt() helper avoids modulo bias by discarding out-of-range draws through rejection sampling so every character or word has an identical selection probability.
When generating a password, the engine runs up to 64 rejection attempts across the combined character pool until the candidate contains at least one character from every ticked class. If the chosen length equals the number of classes, a Fisher-Yates shuffle guarantees uniform placement without favoring earlier positions.
Zero-persistence session history
Clicking Generate new moves the previous value into a collapsible Previous list capped at five entries in volatile component memory. Nothing is written to localStorage, sessionStorage, cookies, or IndexedDB, and zero network requests leave the tab.
Password vs passphrase security comparison and entropy math
Random passwords achieve high security in compact strings by drawing from up to 86 distinct characters, whereas a Diceware-style passphrase generator achieves comparable unpredictability by chaining multiple dictionary words. Both approaches are measured in bits of entropy so you can compare a 16-character string directly against a 7-word phrase.
The NIST SP 800-63B Digital Identity Guidelines emphasize overall secret length and unpredictability over arbitrary complexity rules, requiring verifiers to support secrets of at least 64 characters including spaces. When preparing test fixtures or formatting sample strings, you can also handle case converter privately on your device without exposing sensitive text.
| Mode | Default setting | Pool or wordlist size | Calculated entropy | Strength rating |
|---|---|---|---|---|
| Password (all 4 sets) | 16 characters | 86 characters | 103 bits | Very strong |
| Password (avoid look-alikes) | 16 characters | 76 characters | 100 bits | Very strong |
| Password (alphanumeric only) | 12 characters | 62 characters | 71 bits | Strong |
| Passphrase (with number) | 7 words + 1 digit | 2,394 words | 85 bits | Strong |
| Passphrase (words only) | 5 words | 2,394 words | 56 bits | Fair |
Character pool size and length scaling
Password entropy equals length * log2(alphabetSize), where lowercase (26), uppercase (26), digits (10), and symbols (24) combine to 86 characters. At the default 16 characters, entropy reaches roughly 103 bits, pushing brute-force guessing at 10 billion tries per second to longer than the age of the universe.
Lowering the Length slider to 6 characters drops the rating into Weak (under 45 bits) or Fair (45 to 64 bits) depending on how many character sets remain checked. Keeping all four classes unchecked triggers a reminder because at least one set is required.
2,394-word passphrase entropy calculations
Passphrase mode draws independently from a curated list of 2,394 clean English words between three and eight letters long, yielding about 11.22 bits per word. At the default Words count of 7 with Add a number enabled, total entropy reaches roughly 85 bits (7 * log2(2394) + log2(70)).
Character classes, look-alike filtering, and passphrase separators
Fine-grained toggles in the options sidebar adapt every generated secret to strict form validators or awkward keyboards. Password mode offers independent checkboxes for Lower case (a–z), Upper case (A–Z), Numbers (0–9), and Symbols (!@#$…) alongside a dedicated Avoid look-alikes (I, l, 1, O, 0) filter.
The OWASP Authentication Cheat Sheet recommends storing long, unique secrets in a dedicated password manager rather than reusing memorable variations across sites. Before emailing sensitive documents protected by a fresh key, you can lock them locally with Protect PDF using the exact string you just generated.
Tip: Enable Avoid look-alikes (I, l, 1, O, 0) whenever you must type a password manually onto a smart TV, BIOS prompt, or mobile phone where copy and paste is unavailable.
Excluding ambiguous glyphs for manual entry
Ticking Avoid look-alikes strips ten visually confusable characters (I, l, 1, O, 0, o, |, `, ', and ") from the active pools before sampling. That reduces the full four-class alphabet from 86 to 76 characters, costing only about 3 bits at 16 characters while eliminating transcription errors.
Choosing word separators and optional digits
In Passphrase mode, Between words accepts a hyphen (-), space, dot (.), underscore (_), or no separator. Capitalize words and Add a number default to checked, capitalizing each word and appending a random digit (0–9) to one randomly chosen word so the phrase satisfies legacy site validators that demand uppercase letters and numerals.
When should you use a passphrase instead of a random string?
Use a passphrase whenever you expect to type, memorize, or verbally dictate a credential, such as a device unlock key or a shared Wi-Fi password. Use a dense 16-to-64-character random password whenever your password manager autofills the field or a legacy system restricts spaces and total character length.
Passphrases trade a few extra characters of width for clear human readability without sacrificing cryptographic strength once you select 6 to 10 words. Selecting 7 words with Capitalize words and Add a number enabled reaches 85 bits of entropy, which resists offline brute-force attacks while remaining easy to read across two screens.
You can clean up documentation with strip markdown or generate fresh credentials for a new service locally, knowing every secret stays inside your browser from the moment it is created. Keeping generation client-side removes third-party logging risks completely.