# STEALTH: Steganographic Text Encoding Agnostic to Low-Level Tensor Hardware

DevFeed: [STEALTH: Steganographic Text Encoding Agnostic to Low-Level Tensor Hardware](<https://devfeed.tech/articles/stealth-steganographic-text-encoding-agnostic-to-low-level-tensor-hardware-132291.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16928>)

Author: William Fleshman

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [tokenizers](<https://devfeed.tech/topics/tokenizers.md>)

Tags: [architecture](<https://devfeed.tech/tags/architecture.md>), [code](<https://devfeed.tech/tags/code.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [decode](<https://devfeed.tech/tags/decode.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [encoding](<https://devfeed.tech/tags/encoding.md>), [family](<https://devfeed.tech/tags/family.md>), [floating-point](<https://devfeed.tech/tags/floating-point.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-architecture](<https://devfeed.tech/tags/gpu-architecture.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [key](<https://devfeed.tech/tags/key.md>), [language](<https://devfeed.tech/tags/language.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [low-level](<https://devfeed.tech/tags/low-level.md>), [message](<https://devfeed.tech/tags/message.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [noise](<https://devfeed.tech/tags/noise.md>), [text](<https://devfeed.tech/tags/text.md>)

## AI overview

Bit-exact linguistic steganography can fail when sender and receiver compute different token probabilities, including differences caused by precision, quantization, GPU architecture, or scoring tokens in parallel. STEALTH addresses this by sampling tokens with an exponential race driven by keyed pseudorandom noise; probability mismatches change a token only when the race is close. A rateless spinal code, decoded with beam search, is designed to recover the message despite such flips, without requiring the parties to agree on one token distribution. The authors prove stegotexts remain computationally indistinguishable from honest model samples. In reported experiments, STEALTH decoded every tested message across receiver settings within the sender's model family and across different model families with separate vocabularies and tokenizers. The authors also report higher payload per token than prior robust schemes and no successful distinction by trained steganalyzers or a targeted statistical test.

## Source excerpt

arXiv:2611.16928v1 Announce Type: cross Abstract: Provably secure linguistic steganography hides a message inside text sampled from a language model, so that without the key, a stegotext is computationally indistinguishable from an honest model sample. Nearly every such scheme assumes the receiver can recompute the sender's token probabilities exactly bit for bit. We show this assumption is fragile: across three model families, bit-exact schemes decode almost none of our messages once the receiver's floating-point precision, quantization, or GPU architecture differs from the sender's, or even when the receiver merely scores a text in parallel versus one token at a time, with the same model, precision, and hardware held fixed. We present STEALTH, a scheme built for exactly this setting. The sender draws each token with an exponential race driven by keyed pseudorandom noise, so a receiver's probability mismatch flips a token only when the race is close; a rateless spinal code, decoded by beam search, absorbs the resulting flips without the sender and receiver ever agreeing on a single distribution. We prove stegotexts remain indistinguishable from honest samples under this construction. Empirically, STEALTH decodes every message we sent for every receiver in the sender's model family, at every precision and quantization level, on smaller models, and on a different GPU architecture. STEALTH is even robust across different model families with separate vocabularies and tokenizers, decoding every message we tested. Our approach carries far more payload per token than prior robust schemes, and trained steganalyzers and a targeted statistical test both fail to distinguish its stegotexts from honest model output.