3

I was wondering if it was possible to make any sense from an unknown language by using cryptanalysis. My premises are: suppose an analyst that knows nothing about a language (let's suppose chinese but it can be any other as long as it totally unknown to the analyst), but you do know it's a language (in order to avoid responses like codex seraphinianus) and you can have as much samples of the language as you want (imagine literature, tweets, newspapers, ...)

EDIT: This question came to my mind because I read some time ago about how Americans in WWII used navajo speakers to carry coded messages, hence my interest in trying to solve the problem by cryptanalysis. Also, I'm aware that navajo isn't a written language, but I wanted to assume it in my example for simplicity

mrbolichi
  • 141
  • 4

3 Answers3

3

If you've got many examples of a language then you can probably perform frequency analysis to distinguish correct plaintext from invalid plaintext. This would make it easier to find out if you have successfully broken a cipher.

Understanding the language itself is of course off topic here - language studies / linguistics different from cryptanalysis. But if you have a newspaper then you can put it next to another newspaper with a language you know. You'd certainly find many articles that would have the same content, so you can use these newspapers as a Rosetta stone and "decipher" (most of) the language.

AFTER THE EDIT: Yes, you could certainly perform - for instance - frequency analysis to establish which language constructs are used often and where. This can help deciphering the language. But I presume that it will only result in statistical knowledge of the text.

Finding - for instance - the meaning of nouns will not be helped too much with this kind of analysis, you need to place the language structures in context for that.

Maarten Bodewes
  • 92,551
  • 13
  • 161
  • 313
3

I think the fairest answer that can be given to this question is that cryptanalysts really aren't linguists. To the extent that cryptanalysts deal with language, they do it in a very practical, not theoretical fashion:

  • When cryptanalysts deal with language they look at it through the lens of writing. Linguists on the other hand are trained to think of speech and signs as real language, of which writing is only an impoverished representation.
  • Linguists are fundamentally interested in precise descriptions of language's grammars, and in the grammatical theory of natural languages, as an end to itself. Cryptanalysts really are only concerned with opportunistically spotting redundancies that allow them to zero in on likely plaintexts and steer them clear from hopeless gibberish.

Apart from very generic resemblances (e.g., both cryptanalysis and linguistic fieldwork are iterative processes of collecting data, spotting patterns, formulating hypotheses and testing those with more data), cryptanalysis doesn't equip you to do linguistics. So if "making sense" of language means formulating and testing hypotheses about its grammar or lexicon, then cryptanalysis is not really any help.

If on the other hand "making sense" of a language is just finding regularities to help cryptanalysis of communications likely to be in that language, then linguistics is not particularly relevant. The analyst's task there is to opportunistically find and exploit whatever patterns may be of use, not to explain how the language works.

Luis Casillas
  • 14,468
  • 2
  • 31
  • 53
2

Very interesting question. Let me try and explain this as best I can:

Changing the language is an example of "encoding" which is vastly different from "encryption".

A frequency analysis might help in mapping the most frequently occurring letters (in a descending order), but because all languages have a defined syntax and structure (loosely at least)it would be easy to look this data up which then would push it to being a question of translation.

The essential factor in a crypto service is that the plain text and cipher text look nothing alike, and a desirable feature is that the same plaintext blocks should not yield the same cipher text blocks (when used with a different key, or CTR mode, or Initialization Vectors).

So ideally if you had the phrase "hello world" 1) AES 256 bit (key = 12345) -> qE7HkNCab8p3Mu+FPN5By7WjGLDqG3vfgzCwKGETzEg=

2) AES 256 bit (key = 12344) -> fqWfX6RLvbGZfxH1i1aeSjPxLg1aPIiAkyVOfVrhd9o=

a single change in the key results in vastly different cipher texts.

In an encoding scheme, (such as utf-8), the data is preserved but just translated into another language for a particular compiler or platform to parse.

Whereas when you encrypt, the cipher text is gibberish. It SHOULD not be readable in any format unless the secret key is used to decipher the text.

TL;DR translation (encoding) still retains meaning in some human/computer format.

Encryption retains information, but is essentially gibberish unless the (presumably) secret key is applied to it.

Hope this helps.

baconSoda
  • 131
  • 3