LLM inference with deterministic PRNG sampling enables encoding and recovering 32-bit secrets in the seed from generated text, achieving high accuracy in both known-prompt and unknown-prompt settings.
Provably secure generative linguistic steganography.CoRR, abs/2106.02011, 2021
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Steganography Without Modification: Hidden Communication via LLM Seeds
LLM inference with deterministic PRNG sampling enables encoding and recovering 32-bit secrets in the seed from generated text, achieving high accuracy in both known-prompt and unknown-prompt settings.