[Zenst]

Certainly seems that way. How I read "Noticing some repeated pairs of symbols - which represented letters - the first word cracked by GCHQ boffins was Sidebottom's favourite word, "bobbins"."

Though two triangles are used per letter - you can check that with one of the examples which has a message "Why does my nose hurt after concerts?" - 37 characters in total (including spaces), then count the triangles - 74. Hence two inside triangles are used per character.

But certainly highlights how adding noise to any encryption has it's upsides.

[otwo3]

> Adding noise to any encryption has it's upsides

That's assuming that the secret is the encryption algorithm itself rather than the key. Modern symmetric encryption does not work that way - the algorithm is public and well known while the key is the actual secret required for encryption/decryption.

I don't see how adding noise in modern encryption can help other than increase the size of the output.

Some modes of operation make use of random noise (IV in CBC, nonce in CTR, etc) because it's a convenient way to get a unique number but it's not for obscurity, it's because it's needed to prevent attacks on these modes.

[moomin]

Look up “confounders”, random noise can be extremely useful if you encrypt it as well. This significantly increases the work required to decrypt (since you’ve got to decrypt random noise as well as signal), makes it much harder to tell if you’ve actually decrypted something successfully (depending on how well you can test the plaintext, obviously) and frustrate correlation attacks because every message has a different payload even if the logical payload was the same.