17
$\begingroup$

I have long had this idea of a simple encryption that can be used during spoken conversations to communicate with someone without being understood by other people in the room. Is there anything like that?

The criteria are as follows:

  • Does not have to be secure enough withstand anyone trying to crack it using a recording or other technical helpers
  • Does need to be secure enough so that people cannot understand it on the fly
  • Needs to be key-based so that there are enough keys that you wouldn't understand it without knowing the key, even if you know the encryption method
  • There need to be enough possible keys so that you can't just guess the key
  • Needs to be easy enough to encode/decode that you can do it in slow conversation speed if you trained enough for it

Is there anything like that? If not, is it at all possible?

$\endgroup$
  • 3
    $\begingroup$ There are a lot of "scrambling" techniques similar to Pig Latin, but more complex. It isn't key-based though. $\endgroup$ – forest Jul 25 at 8:08
  • 2
    $\begingroup$ I used heavy slang with a friend one vacation to avoid being listened to by english speakers without deep cultural understanding. Instead of asking him "are you ready to go?", I would quietly and quickly say something like "ya finna bounce?" instead. "Yes" becomes "fa shizzle" or "pope wear a funny hat?". Words like "him" become "dat mofo", etc. Worked really well and was a lot of fun without too much thinking. You might also look into eggy-peggy, but it takes practice. $\endgroup$ – dandavis Jul 25 at 20:06
  • 6
    $\begingroup$ How about learning a foreign language? :) There are about 6500 spoken languages in the world apparently. $\endgroup$ – Federico Poloni Jul 26 at 7:18
  • $\begingroup$ How about sign language (as used by baseball coaches)? $\endgroup$ – Hagen von Eitzen Jul 27 at 23:26

11 Answers 11

22
$\begingroup$

We're not computers. We'd have to do calculations in our heads while talking about something other. There are quite a few parts of the brain active when we're talking, it would be tricky at best to put something in between.

There is the notion of code talkers but those used their own language such as some of the Native American languages to communicate. Such a pre-learned language however doesn't have a key so it doesn't fulfil your requirements completely.

As forest mentions in a comment, there are quite a few ways of scrambling techniques such as Pig Latin that act on the language used. Without a (sufficiently large) key those should be thought of as obfuscation techniques rather than encryption techniques, even if they are effective for their purpose.

Simple substitution of words could also work, but without practice for encoding / decoding using a particular key I suppose this will be hard to do as well. It might be possible if you'd use an extremely limited vocabulary (the "alphabet" in the cryptographic lingo) and message size such as just "yes" and "no".


Note that I've not gone into techniques of even hiding that you're using encryption. That would be seen as oral steganography I suppose. Or, if you take it one step further, deniable encryption. Without feasible oral encryption I guess these are mostly theoretical concepts.

$\endgroup$
  • 2
    $\begingroup$ For the philosophers out there: we might actually be computers, but rather highly complex and faulty ones - we're not to be trusted with any calculation that is supposed to be deterministic :) $\endgroup$ – Maarten - reinstate Monica Jul 25 at 8:02
  • 15
    $\begingroup$ We're bio-optimized prediction engines, which tend not to do very well in abstract mathematics. $\endgroup$ – forest Jul 25 at 8:09
  • 1
    $\begingroup$ A more "advanced" version of Pig Latin is Polari - a language consisting of mixture of a lot of other languages like Romani, Yiddish, and others, as well as some slang versions of each. It was used in public spaces by people who wanted to conceal what they are talking about. For example, gay people were still prosecuted at the time so a usual practice was to drop a few words in Polari to see if another party picked it up. If they didn't, they likely weren't gay, so it was a safe "test" to look for partners. $\endgroup$ – VLAZ Jul 26 at 13:59
  • 1
    $\begingroup$ I seem to remember reading that some pairs of Navajo code talkers had pre-arranged codes for scrambling numbers, so that 2343 yards might be spoken as 7121 yards. This constitutes a rudimentary key of some kind. $\endgroup$ – TonyK Jul 26 at 21:22
10
$\begingroup$

The one thing no one else has touched on is speed. Any cipher that you have to think about consciously will be too slow to be practical, and too easy to break to be effective. It would also be vulnerable to an eavesdropper writing down on pen and paper, and then cracking it later (not sure if that counts as recording).

Ciphers like Rövarspråket and the B-language from Bernd Wilke work by just increasing the amount of sounds per time. They are really easy to crack if you can take your time to decipher them, but are impossible if spoken quickly. There's simply too much information to process, so your subconscious brain drops parts of it, and what's left for the conscious brain is too destroyed to be of use. Without any recording allowed, they are therefore secure enough for this threat model, even if pen and paper is allowed.

In signal processing terms, this is roughly equivalent to Direct-sequence spread spectrum.

So you and your friend would generate a mapping from phonemes to groups of phonemes (the key) and practice it beforehand. When used with others nearby, even if they figure out the key, they have no chance of cracking (or writing down) anything you say until they've practiced it for a while, giving you long enough to switch keys.

$\endgroup$
  • 1
    $\begingroup$ I think speed and pitch both could be used. People are absolutely able to modulate these things on the fly, and we've really gained a nack for hearing them over the millennia. I still don't think anyone in this Q&A has come up with a really truly approachable way to use them, but it's for sure got to make some use of these parallel forms of expression. $\endgroup$ – Seph Reed Jul 27 at 9:16
4
$\begingroup$

One form of spoken encryption that would be strange, and somewhat breakable, but doable is a sort of sarcasm/singing mixture. If two people know the exact same song, they could sing each word. Then, if any word is sung at the wrong note, it would imply "opposite."

It kind of works off the basis of sarcasm being like an encryption, where someone who is especially deadpan will make it so you have to know some outside data to interpret whether they're saying one thing or another.

Most people aren't that deadpan and give it away with a joking intonation, which is sort of a well known key. But by switching the key to something only two people could know, you'd have to know the answer to details about what they're talking about to tell whether any given statement was true, or anti-true.

You'd still be able to tell largely what they were talking about, but it's a start.

EDIT: Also, given how many different tones are acceptable (sound good) in a song at any given point, it wouldn't be hard to choose off notes without sounding bad. It's called "substitution" in jazz improv.

$\endgroup$
  • $\begingroup$ So the key is to sing out of key? That would be a single bit key in that case. You could use notes, but first of all the key would be transmitted while sung. So it doesn't comply with Kerkhoffs principle. And if you make it more complex than that then it will likely fail. And you wouldn't learn it to my family, we're not tone deaf, but close enough. And our own songs would always sound "false". Not a critique on the answer - interesting enough - but I'm not sure about the practicality of it. $\endgroup$ – Maarten - reinstate Monica Jul 25 at 19:25
  • 1
    $\begingroup$ You don't sing out of key. I just made the edit before this comment came. At any given point, there will be 4 other very easy tones to choose (pentatonic is childrens nursery rhymes). If you can sing mary had a little lamb, you can do this. $\endgroup$ – Seph Reed Jul 25 at 19:29
  • $\begingroup$ Also, the key would only be transmitted if someone sang every single note right, and even then an observer would not be able to tell. $\endgroup$ – Seph Reed Jul 25 at 19:30
  • $\begingroup$ If you want to test your ability to do this in real time, try singing twinkle twinkle, while moving up in pitch for "star". You'll hit a note, it'll be wrong, but if someone didn't know the song, they wouldn't be able to tell. $\endgroup$ – Seph Reed Jul 25 at 19:32
  • 3
    $\begingroup$ All I can image here is Dori speaking whale in "Finding Nemo" $\endgroup$ – b degnan Jul 26 at 1:18
3
$\begingroup$

I like the idea that Seph Reed posted in another answer about using songs for it, though not the part about using the tone.

There are several possibilities for using songs as encryption/obfuscation, though they might not be easy to adapt to use keys:

  1. Word substitution: You pick a song and randomly change some of the words to your message - not consecutively, but e.g. "All we are is dust in the wind" becomes "All I am is dust in the home" to spell "I am home". The chosen song becomes your key and you need shared knowledge of the song to decode it. The more obscure the song, the stronger the key, because any outside observer would need to know the song to know which words were replaced.
  2. Title "lookup": You sing/speak lines of a song and the first, last or any other prearranged word of the title is the word you encode. "Lunacy has found me. My eyes seek reality." become "Battery" and "Low Man's lyric" from Metallica, so decode to "Battery low". This requires both to know a lot of songs, but can't be deciphered without having the same knowledge. You can use the word index in the title as a key or use a specific genre, e.g. Jazz.
  3. Tamarian twist: Named for the famous star trek episode. You say the title of a song or sing/speak a line of it and a word or phrase you both connect with the song is the code. For example, if you met when the famously misunderstood Bruce Springsteen song was playing, "Born in the USA" could be code for "meet me". This requires a shared understanding of the meaning of songs and is hard to decipher for others, because the code could be almost completely arbitrary. As a key you could use a line index, e.g. "always sing the fifth line of the song".
$\endgroup$
  • 1
    $\begingroup$ I definitely understand the desire to move towards words, but I think it makes the key easier to break from the outside, and harder to pass key as well as interpret something live. I don't think my solution is quite complete, but I strongly believe moving away from tone is a regression. It's the only ordered yet arbitrary thing that couples to language without changing its meaning that I know of. It's even called singing in key. $\endgroup$ – Seph Reed Jul 26 at 18:03
2
$\begingroup$

One potential solution that covers your first 4 points well and the last point slightly would be to map English syllables to other English syllables using a predefined encryption method and key.

For example Hello becomes H-el-o and encrypts to ay-de-in.

Whilst this would be impossible to do in the moment with any assistance this new encrypted version of language could easily (maybe!) be pre learned prior to a conversation and could definitely be learned for a set amount of phrases. Anyone reading this would already have learned hello's encrypted form. Anyone listening even if they were recording would need to know the key to decrypt the speech.

If the speech was short enough to be remembered the listener could go away and decrypt it themselves even if they hadn't learnt the encrypted language as long as they knew the encryption process and key. Or if you precalculated the conversions and had both speaker and listener had a lookup guide to compare against they would be able to slowly converse.

$\endgroup$
1
$\begingroup$

Use a simple monoalphabetic substitution cipher. To maximize the chance of getting pronounceable cipher words, choose a code where vowels map to vowels and consonants map to consonants, leaving 5!×21! ≈ 6.13×1021 possible keys.

Come up with whatever mnemonic devices you need to remember the code. For example, for the mappings D→Y, F→K, and G→P, think of the words “day”, “fork”, and “gap”.

Of course, this method will have all the cryptographic weaknesses of any monoalphabetic substitution cipher, but it should be perfectly usable without a computer.

$\endgroup$
1
$\begingroup$

I don't think anything requiring the user to do mental calculation on the fly is going to be terribly reliable, especially under stress.

I think something like a grill cipher might be doable: Embed the words from the desired message in a rambling sentence with a lot of extraneous words, then identify the words that should be kept by some mean. Blinking during legitimate words from the message text would be simplest, but you could also memorize a numerical key and have the digits of the key determine how many words of nonce there will be until the next legitimate word. (A lot of science wonks I know have memorized pi to a ridiculous number of digits, so that might do.)

$\endgroup$
1
$\begingroup$

Karate practitioners practice kata -- formalized sequences of 30-60 moves or so, including steps, stances, blocks and strikes.

Since they spend a lot of time practicing these, they get built into muscle memory, and they can perform the sequences, or visualize them, without thinking.

If you and I were both Karateka, then we would use these kata as keys in a spoken code. If we agree on "Heian Godan, right foot" as a key, for example, then when either of us talks, we together visualize performing the Heian Godan kata repeatedly, and we'd both understand that only the words spoken when the right foot moves actually count. Other words would be misdirection.

If we were advanced members of the same dojo, we could have up to 20 katas or so to choose from, which is a pretty good key space.

If you speak in the rythm of the Kata, you get automatic correction for synchronization errors, too, and you'll sound like Captain Kirk :-)

Lots of martial arts and other practices have katas or other kinds of complex rituals that would work.

$\endgroup$
0
$\begingroup$

Your question reminds me to a language we used as children. although I don't know if it really fulfils your requirements.
And I don't know if this is possible in english at all, maybe it is a German special. In German you pronounce letters nearly always the same, especialy the vowels.

The B-language:

You replace each vowel in the words with the vowel,a 'b' and the vowel again.

Trying to give some English examples, respecting only spoken vowels:
my house has a red roof. -> myby houbouse habas aba rebed rooboof.
I like the weather in spain-> Ibi libike thebe weabeatheber ibin spaibain
it could be complicated if you use longer words -> ibit coubould bebe cobomplibicabatebed ibif youbou ubuse lobongeber wobords

Remember to speak the repeated vowel like the origin vowel.

The key could be the consonant you insert, maybe you can even use syllables.
Variation: instead of vocal repetition only a syllable appendage

$\endgroup$
  • 4
    $\begingroup$ I don't think this meets the criteria in the question, specifically "needs to be key-based so that there are enough keys that you wouldn't understand it without knowing the key, even if you know the encryption method" and "there need to be enough possible keys so that you can't just guess the key". If you know how it works, it's pretty easy to tell which letter has been inserted. $\endgroup$ – Ilmari Karonen Jul 26 at 10:16
0
$\begingroup$

The problem with any obfuscation technique is that it has to be simple enough for Alice and Bob to make the transformations necessary, but not so simple that Eve can trivially do the same.

The closest thing I can think of is Pig Latin,

Pig Latin is an obfuscated form of English, which is translated into Pig Latin word by word. To translate an English word into Pig Latin, the beginning consonant or consonant cluster is shifted from the beginning of the word to the end of the word, and following it the vowel sound "ay" is added.

For instance, using the common rules

Ix-nay on the upidstay

de-obfuscates to

Nix (reject) the stupid

The key, simple though it is, is just the specific pre agreed set of rules used in the transform.

$\endgroup$
0
$\begingroup$

Spoken encryption - Is there anything like that? If not, is it at all possible?

Requirements: Key based.

The key is essentially prearranged by rehearsing/learning, without that one would ask for clairification (which stymies the flow). There will always be some people whom readily understand, much as there are good cryptologists, and some whom won't retain the information, such as poor listeners.

There already exist means to codify a language:

  • Vernacular and usage

    People from a common background (area, school, age) use register (leaving out letters), dialect (variety, social class or ethnicity), or sociolect (social or economic class).

  • In-crowd

    People from a select group whom understand what is important to them and what the varying levels of significance of each word used is to the sentence in which it is used.

  • Big-words (italicized because the words are usually not big, just a group of small words)

    People who use an occasional word that isn't understood by one other listener or string together a few words to create a new meaning are often accused (sometimes jokingly) of using big words.

  • Internet slang and SMS shortcuts or Chat abbreviations

    Peopla usin' mour TLAs, OMG.

$\endgroup$

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