# Any benefit to writing a message as a mix of 2 languages before encryption?

For example, imagine a message that is a mix of Japanese and English words; Japanese in unicode, English in ASCII. Japanese doesn't use spaces between words.

The sentence structures, anything abstract, and verbs would have to be in Japanese. Anything directly translatable such as colors, transportation machines, proper nouns, etc. would be in English.

Would writing the original message as a mix of 2 languages (especially a Western and non-Western language) contribute to making it more difficult to decrypt?

Historically, there did exist a benefit to using a language that the adversary was not familiar with. The name for this is code talkers, and the most famous ones (at least in the USA) are the Navajo code talkers of World War II. The idea was to defeat attacks that relied on statistics about the language used in the plaintext.

In modern cryptography, however, there is no benefit, because modern ciphers are designed to obscure any patterns in the plaintext—even patterns that we haven't thought of. This is because modern ciphers are expected to be not just "hard to crack," but indistinguishable from random data.

There's a second reason why using multiple languages would not have any benefit, which is that you're using a computer, so no matter what language you use you're still encoding text into a character encoding like UTF-8—and this already gives you statistical patterns that an attacker could look for.

Third reason: English and Japanese are both widely spoken languages with lots of example text available. There's a reason why code talkers tended to use aboriginal languages.

Fourth reason: Kerckhoffs' principle. Your cryptographic system should not rely on the secrecy of simple tricks like what you propose.

• In a practical sense, I mean just reading Japanese / English text when everything is written using English characters takes major mental firepower. No word spacing allows you to write ambiguous text. Without a high-level abstract understanding, it would be impossible to read. You couldn't understand where words breaks are. It'd just be a list of English characters. anyway... thanks for the answer! Jun 22 '16 at 21:37

If you're referring to a classical cipher, it might complicate frequency analysis and other such techniques.

For a modern cipher, it makes no difference. Modern ciphers operate on arbitrary patterns of information. Ideally, the ciphertext of a modern cipher should have no relation of any kind to the associated plaintext, other then the key.

• Couldn't think of anything to cite in the above answer... Normally I don't answer unless I can blue link to back it up, but I'm not sure what I would link to here. If this wasn't informative enough, someone else will come along and help out. Jun 22 '16 at 20:32

Around and about one hundred years ago, your idea would surely have made sense… but nowadays, modern technology and evolved cryptanalytic techniques are too smart to have a real problem coping with something like that. (Also see my related answer to “Why was the Navajo code not broken by the Japanese in WWII?”)

Even when we completely ignore Kerckhoffs’ principle that “the enemy knows the system” (something we usually assume to be true in the realms of professional cryptography) the answer would still be a plain and simple “not really”.

As a matter a fact, you might as well write your whole thing in Navajo and mix in some Klingon before encrypting things… and it still wouldn’t buy you any free lunch in terms of difficult to decrypt.

### Tip: “multiple encryption” is a more solid alternative in terms of security.

If you really want to add some decent security, you might want to know that there is a much more constructive way to handle things: use multiple ciphers – practically encrypting your plaintext two or more times using different algorithms and/or different cipher modes.

Some (not all!) random examples to give you an idea of what I mean:

• using multiple ciphers:
plaintext $\rightarrow$ encrypt with AES block cipher $\rightarrow$ encrypt with ChaCha20 stream cipher
• or, using one cipher but multiple cipher modes of operation:
plaintext $\rightarrow$ encrypt with AES in GCM mode $\rightarrow$ encrypt with AES in CTR mode
• or, maybe even a combination of the above:
plaintext $\rightarrow$ encrypt with AES-CGM $\rightarrow$ encrypt with ChaCha20 $\rightarrow$ encrypt with AES-CTR

You could go crazy with this until you run out of time, resources, and/or combinations of modes and algorithms… but usually, multi-encrypting your plaintext with 2 different ciphers is more than enough.

As long as you avoid broken things like RC4 and as long as you encrypt things correctly (meaning: “by the book”, thereby avoiding stepping into cryptographic pitfalls), multiple encryption will definitely be more worth the effort as it will make things so much more difficult to decrypt than any language-mixing you might attempt to do on your plaintext message before actually encrypting it.