Compression + encryption can be dangerous if the compression rate is exposed somehow (between messages or within packets of a message).

> we show that it is possible to identify the phrases spoken within encrypted VoIP calls when the audio is encoded using variable bit rate codecs

https://crypto.stackexchange.com/a/2188

See also https://breachattack.com/ when the plaintext is partially attacker-controlled.

If nothing else it would make a great twist in a fiction setting.

These paraphrasing instructions could be followed. But the paraphrasing could be done using some LLM. A sufficiently advanced adversary manages to invert the model somehow, and as a result can get the original plain text out of the paraphrased message, which lets them do a known-plaintext attack, get the key, and use it on other messages.

Sort of technobabble (is the idea of inverting an LLM nonsense?) but fun.

You add a random number to the encryption key, and also send that random number (seed) as part of the message.

Boiled down to the very essence modern cryptography is: Using a secret seed plus a public seed, generate a long random number (of the same length as the message), then XOR that number with the message.

The hard part is generating that random number in such a way that you can not reverse the process and reclaim the secret seed.

Lookup "initialization vector" for more.

In short: cipher modes, IVs, nonces, and random padding.

Generally speaking, in two different ways: (a) cipher modes will usually use a combination of initialization vectors, block chaining, or an incrementing counter to perturb the encipherment of each block so that repeated data does not result in repeated ciphertext and (b) encrypted protocols will include a section to be filled with random "nonce" data so that repeatedly enciphering the same message will also result in different ciphertext, and might also add random padding so that the length of ciphertext can't be used to deduce the length of a particular message.

Modern cryptography solves this by using randomness (IVs, nonces, padding, salts) so that even identical plaintexts encrypt to different ciphertexts, eliminating predictable patterns.