Generated password characters that are hard to mistake

Question

I am developing a tool where a short pass code is generated. That code will be read to the intended user, so I want to limit my character set to letters that are hard to mistake. Both the reader and the listener will be untrained, so any solution must be transparent to them. I want to be able to just give them the passcode and have it exceedingly unlikely that a letter will be misheard.

First off I have limited my possible character list to those characters that look the similar depending on font, so no l or 1, 0 and O are out, as are S and 5, any others you can think of?

Also, I want to remove letters that sound similar, so G&D and B&P are out. Is there a resource of commonly misheard letters?

I know this will cut down on the possible character count severely, but I do need it because of accessibility issues. Also, the person on the receiving end of the code probably won't be aware that I am limiting the character set. The fact that I am limiting the characters should be transparent to both the person reading the password and the person hearing it.

Answer 1

Can you not have it read back with the NATO phonetic alphabet?

A Alpha B Bravo C Charlie D Delta E Echo F Foxtrot G Golf H Hotel I India J Juliet K Kilo L Lima M Mike N November O Oscar P Papa Q Quebec R Romeo S Sierra T Tango U Uniform V Victor W Whiskey X X-ray Y Yankee Z Zulu

If not, you really need to drop any letter with an "ee" sound at the end, i.e.: B,C,D,E,G,P,T,V,Z

Answer 2

I found a similar discussion on google answer: http://answers.google.com/answers/threadview/id/521339.html

It might help you out, check out the answer given by pinkfreud-ga

Answer 3

Hmm... I'm assuming that the code will be a visual representation that will then be read to the end user? because if it was spoken by speech synthesis, then 0 and O wouldn't be an issue, right?

If so, would you be able to visually aid the representation of that code?

Your code is
[M] ike [C] harlie [5] five ...

Answer 4

This approach seems a little harder than it needs to be, tbh, especially with the numeric similar appearance limitation. Why not restrict letters and numbers to certain positions within the passcode, and make it clear that "the first X characters will be alphabetical characters, and the last X characters will be numeric"?

The NATO phonetic alphabet like JNKyle says would also be helpful for visual feedback if it's possible for you to use it.

Answer 5

Pick only few characters and use those:

APJZCHWRTE
apjzchwrte

Though a good password ends up being long. add few digits and you can construct shorter passwords.

Answer 6

Setup an easily readable charset and then randomize it.

Here's a solution in Ruby

charset = %w{ 2 4 6 7 f h q w x y }
password = (0...10).map{ charset.to_a[rand(charset.size)] }.join

This produces passwords like ""qh4w26wxfq"

Answer 7

Some out of the box options...

If it doesn't have to be random numbers and letters, you could go with a dictionary of words. Then the person would not have to read off each character. The word could then be padded with numbers to make it more unique.

3power41
69topic31

Another possibility, make it all sets of two characters adjacent to each other.

ab-pq-78-de-mn-op

It would be easier to notice when a character breaks from this pattern because it was misheard (even someone untrained would probably notice the pattern).

If typed, a similar pattern could be used as adjacent letters on the keyboard.

qw-df-hj-io-23-we

The possibilities are limitless.

Answer 8

After taking the suggestions in other answers I this is the reduced list of characters that I have found to be hard to mistake when spoken or viewed.

fhikorsuwxyFHKLQRUWXY246789

Special thanks to:

• David Hedlund
• Kieren Johnstone
• JNK
• Lobsterm

Without their suggestions I would't have been able to synthesize this list.