3 ms·
Correct, but the recommendation in the article is still wrong. If I get a password that has four Unicode characters in it, it's possible that it's a strong pass
by lolinder 4y ago
Correct, but the recommendation in the article is still wrong. If I get a password that has four Unicode characters in it, it's possible that it's a strong password that chose truly randomly from the available Unicode characters, but it's much more likely that someone just picked four that they like. In the latter case, the bits per character is going to be much lower, but I can't tell the difference just looking at the password.
The only good solution I can think of is to assume low bits per character and require a minimum password length.
- zinekeller 4y ago> but it's much more likely that someone just picked four that they like Or as I would call it, the fallacy of uniformity. The author assumes that hackers would only try a dictionary attack using lower case letters, but this is no longer true since mid-2010s. Some dictionaries geared for international services also have non-Latin characters since not all countries restrict passwords to alphanumerics.
- alanfranz 4y agoIf it’s four random Unicode code points, it’s not a dictionary word.
- Aeolun 4y agoYou don’t have to assume this. You can check the length in bytes of the password.
- alanfranz 4y agoThe same problem can happen with eight-char passwords. I could pick the same eight chars I like all the time and reuse passwords. You cannot fix password reuse that way. I'm not advocating for emoji passwords, it was just a discussion about password strength vs length.
- lolinder 4y agoRight, you can't force people to use good passwords, but you can set up requirements that try to maximize the chances that that's what you get. As others have mentioned, there's no way to fully evaluate password strength short of providing your own entropy, so while the distribution is an important one to know about there's not much we can do to apply it while retaining a traditional password model. NIST has some recommendations[0] that can detect when a password is not strong, such as checking a list of cracked passwords, but we can't prove that it is. [0] https://pages.nist.gov/800-63-3/sp800-63b.html#memsecret https://pages.nist.gov/800-63-3/sp800-63b.html#memsecret