Steganeur – Hide secret messages in LLM-generated text (Rust)
Source
github.com
Author
marcsnid
Date
Context
Steganeur hides a message in a LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition →'s choices of text units, or tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →. Its encoder uses the message's bits to select among possible continuations; a recipient reconstructs those choices using matching model settings and context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →. The visible prose becomes the carrier for another message.
Recovery depends on reproducing the encoding scheme. The README describes methods whose decisions shift when model probabilities vary between runs, potentially corrupting the recovered bits. A fixed-hash method avoids that particular dependency, with a trade-off in how natural the prose sounds. The implementation therefore makes determinismWhether the same input reliably produces the same output — something LLM systems mostly lack, which changes how you test and debug them.Full definition → part of the communication problem.
Terms in this piece · Glossary
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
determinism — Whether the same input reliably produces the same output — something LLM systems mostly lack, which changes how you test and debug them.