Disclaimer: I was looking for a place to ask this question on SE and this site seems to be the most fitting. If this question doesn't belong here, feel free to redirect me.

Just out of boredom and curiosity, I've started to develop my own secret notation alphabet (for manual use, no computers involved). It's based on Elian Script, although I scrambled the letters in a non-obvious way (no ROT13 or similar) and expanded on the form a bit. I still feel as if it wouldn't be that hard to "decode" my notations, though.

So my questions would be:

  • what further methods are there for making an alphabet 'safer', except using different letters and scramble them?
  • what methods are there for text-encryption that could be used for manual everyday use? (like leaving out punctuation, for example.)
  • $\begingroup$ What you have is likely a substitution and transposition cipher. To make it harder to decode, for starters you can use a code book (for common bi- and trigrams, perhaps) and/or shift the key by some amount for each character or group. But it'll still almost certainly be trivial to decode for someone who tries. What you'd almost certainly want to do is remove the statistical significance of the various glyphs, which even with transposition can give vital clues as to the content of the message. Leaving out punctuation only works insofar as it obscures the structure of the plaintext. $\endgroup$
    – user
    Sep 19, 2013 at 12:32
  • $\begingroup$ @Michael Kjörling: That's a good start, I'll read up on that and see what I can use without slowing the process of writing considerably. I'd probably write everyday notes with that alphabet, so most methods seem to slow me down so much it's not viable for everyday writing anymore. Although I'm sure that with extensive practice I'll get it up to speed. Anyway, thanks. $\endgroup$
    – Jenny
    Sep 19, 2013 at 13:21
  • $\begingroup$ or learn mirror writing, most would just think that it is written with exceptionally bad handwriting and in some strange language and not think of holding it in front of a mirror $\endgroup$ Sep 19, 2013 at 14:03
  • $\begingroup$ You might enjoy looking at shorthand systems such as Gregg shorthand -- they were never intended to be difficult to decode, but some stenographers use so many of their own custom abbreviations that even other stenographers using the same system have difficulty reading it. $\endgroup$
    – David Cary
    Nov 12, 2013 at 3:16

1 Answer 1


A Michael Kjörling notes, what you basically have is a monographic substitution cipher. These are generally easy to crack using frequence analysis; for example, if we know that the underlying text is in English, we can be fairly sure that the two most frequently occurring symbols correspond to the letters T and E (and that any symbol that occurs unusually often between them is probably H).

To make your cipher stronger against frequency analysis, you want to somehow make the symbol frequencies in your ciphertext as uniform (and as uncorrelated with nearby symbols) as possible. There are many ways to do that, but the requirement that the resulting alphabet should still be something one could learn to write fluently rules out most of them. Still, there are some things you could do:

  • Come up with multiple (preferably completely different-looking) symbols for the most common letters in English. The number of symbols per letter should be approximately proportional to its frequency. Alternate between the symbols randomly, or according to some rule of your choosing, or just as you feel like.

  • Conversely, consider merging some rare letters together under the same symbol(s). This works best for letters that have shared historical roots, such as I / J or U / V / W.

  • Consider making up separate symbols for common or distinctive letter combinations like TH, QU or ING. These should also have multiple variants according to their frequency. Some particularly common words should probably have dedicated symbols.

  • Conversely, consider encoding rarer letters using two symbols, e.g. as the symbol for a more common similar letter plus a modifier symbol (which could be used for multiple letters to make it more common, and which could also have a stand-alone meaning or just be used as "chaff").

  • Introduce lots of stylistic variations to your symbols, so that it's hard for a would-be decipherer to know if two symbols are the same or not. This is most effective if some variations (like line length, or the presence of dots or loops) are essential for distinguishing some symbols but vary arbitrarily for others.

  • Make it hard to tell where one symbol ends and another begins, e.g. by allowing symbols to be combined or written in unusual order.

  • Introduce random meaningless "chaff" symbols into your writing.

  • Make lots of misspellings, either more or less systematically (e.g. randomly interchange S / Z or C / K in words where it doesn't matter, spell some words phonetically or in "txt spk"), or just drop or interject random letters into words where it doesn't introduce ambiguity.

  • Use idiosyncratic jargon, so that even if the text is decrypted, it may still not make sense.

  • If you don't use dedicated symbols for common words like "the", consider substituting them with other words. A substitution like, say, "the" → "cat" could really frustrate a cryptanalyst.

  • Drop unnecessary common words entirely. You don't really need all those articles, do you?

Edit: Last but not least — don't write too much. With enough ciphertext to analyze, any combination of these measures can still be cracked by a clever, skilled and persistent cryptanalyst. Your best defense is making sure that there simply won't be enough text for a reliable analysis to be possible.

Oh, and never write down anything both in cipher and in plain. Even a short bit of known plaintext in the hands of a skilled cryptanalyst can make all these tricks useless. Look up e.g. the Rosetta stone for a historical example.

  • 1
    $\begingroup$ Truth be told, all of these techniques have been tried (historically) to frustrate cryptanalysis... and none of them stood up against cryptanalysts. All of this stuff makes the cipher stronger, but not strong. I wouldn't use any such method for vitally important information. Of course, I think you knew this, but I just wanted to caution others against thinking these methods would greatly improve security. $\endgroup$
    – Reid
    Sep 19, 2013 at 15:57
  • 1
    $\begingroup$ @reid unless you are going to carry an enigma machine around it's the best you are going to get for on the fly encoding/decrypting, in the end the human brain doesn't handle such things well, an other option is adding another language in the mix (code talker style) $\endgroup$ Sep 19, 2013 at 22:53
  • $\begingroup$ That's very helpful, thank you. I'd vote the answer up, but I don't have enough reputation -.- $\endgroup$
    – Jenny
    Sep 20, 2013 at 7:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.