Password vs Passphrase: Which Is Stronger?
Last updated Published by Hyperdusk Ltd
The short answer
Neither is stronger by nature. Strength is the number of equally likely results the generator could have produced. A random 12-character password drawn from the 94 printable ASCII characters carries 78.7 bits of entropy; a random six-word passphrase from the EFF large wordlist carries 77.5. The passphrase is longer but easier to type and remember.
Generate a random password or passphrase →Key differences
| Random password | Random passphrase | |
|---|---|---|
| Built from | Characters picked at random | Words picked at random from a fixed list |
| Entropy per unit | 6.55 bits per character (94 printable ASCII) | 12.92 bits per word (7,776-word list) |
| Length for about 78 bits | 12 characters | 6 words, about 47 characters |
| Length for 128 bits | 20 characters | 10 words, about 79 characters |
| Easy to remember and type | No | Yes |
| Best for | Secrets a password manager stores and fills in | Secrets you type by hand |
What makes a password or a passphrase strong?
Entropy, measured in bits: the base-2 logarithm of the number of results the generator could have produced. In the EFF's words, adding one bit doubles the number of guesses required. A character picked from 94 possibilities adds 6.55 bits. A word picked from a 7,776-word list adds 12.92.
The arithmetic only holds when every choice is made at random. A phrase a person thinks up, such as a quote or a song lyric, has far fewer likely alternatives than its length suggests. NIST makes the same point about passwords in general: people choose secrets that can be easily guessed, and estimating the entropy of a user-chosen password is hard, which is why its guidance is based on length rather than on an entropy estimate.
How long does each need to be?
The table gives the shortest length that reaches a target strength for each kind of random secret. A smaller character set needs more characters; a word list needs fewer units but far more keystrokes.
| Drawn from | 64 bits | 80 bits | 100 bits | 128 bits |
|---|---|---|---|---|
| Digits (10) | 20 | 25 | 31 | 39 |
| Lowercase letters (26) | 14 | 18 | 22 | 28 |
| Letters and digits (62) | 11 | 14 | 17 | 22 |
| Printable ASCII (94) | 10 | 13 | 16 | 20 |
| EFF large wordlist (7,776 words) | 5 words | 7 words | 8 words | 10 words |
How strong is a passphrase by word count?
The EFF large wordlist has 7,776 words, the number of ways five dice can land, so each word can be chosen with dice or by software. The EFF recommends six words from this list for most uses, "for a strength of 77 bits of entropy".
| Words | Bits of entropy | Possible passphrases | Average length | Random password of equal strength |
|---|---|---|---|---|
| 4 | 51.7 | 3.66 × 10^15 | 31 characters | 8 characters |
| 5 | 64.6 | 2.84 × 10^19 | 39 characters | 10 characters |
| 6 | 77.5 | 2.21 × 10^23 | 47 characters | 12 characters |
| 7 | 90.5 | 1.72 × 10^27 | 55 characters | 14 characters |
| 8 | 103.4 | 1.34 × 10^31 | 63 characters | 16 characters |
This page gives no "time to crack" figures. They depend on an assumed guess rate, and that rate depends entirely on how the service stores the password and whether guesses are rate limited.
What does NIST recommend?
NIST SP 800-63B, revision 4, sets rules for the services that check passwords rather than for the people who choose them. A password used as the only authentication factor must be at least 15 characters. One used alongside a second factor may be shorter, with a minimum of eight. Services should accept at least 64 characters, and the guideline says that limit exists to support passphrases.
Services must not impose composition rules such as requiring a mix of character types, and must not require periodic password changes; a change is forced only when there is evidence of compromise. New passwords must be compared against a blocklist of commonly used, expected or compromised values. Services must allow password managers and autofill, and should allow pasting.
On passphrases, the appendix on password strength says that using one "is often an effective way to create a longer password".
When is a passphrase the better choice?
When a person has to type or remember it: a password manager's master password, a disk-encryption passphrase, a device login. Six ordinary words take more keystrokes than 12 random characters, but they are easier to remember.
For everything a password manager fills in, a random password is the better fit. It reaches the same strength in about a quarter of the length, and no one has to remember it.
If typing effort matters most, the EFF also publishes short lists of 1,296 words. Each word adds 10.3 bits, so more words are needed for the same strength.
Which should you use?
Use a random password of 16 characters or more for accounts stored in a password manager. Use a passphrase of at least six random words for the few secrets you type yourself.
In both cases, let a generator make the choices. Either form is much weaker when a person invents it.
Tools for this
- Generate a random password or passphrase →
Password Generator. Makes random passwords of 4 to 64 characters and passphrases of 3 to 10 words from the EFF large wordlist, and shows the entropy of each result. Its full character set has 87 characters, so its figures sit slightly below the 94-character column here.
- Generate a random PIN →
PIN Generator. For numeric codes, where the digits row of the length table applies.
Sources
This page was written from the primary sources below, read on .
- NIST SP 800-63B-4: Digital Identity Guidelines, Authentication and Authenticator Management
NIST. Password length minimums, the ban on composition rules and periodic changes, blocklists, password managers, and Appendix A on password strength.
- EFF's New Wordlists for Random Passphrases
Electronic Frontier Foundation. The 7,776-word large list, 12.9 bits per word, the six-word recommendation and the 1,296-word short lists.
- EFF Dice-Generated Passphrases
Electronic Frontier Foundation. How to pick words with five dice, and the minimum of six words.