ReTokSync: Self-Synchronizing Tokenization Disambiguation for Generative Linguistic Steganography

2026-04-28

AI Summary

This work addresses the practical challenge in generative linguistic steganography where tokenization ambiguity between sender and receiver often leads to desynchronization and extraction failure. To resolve this, the authors propose a self-synchronized disambiguation framework that continuously monitors the receiver’s tokenization during generation and triggers localized corrections only when ambiguity is detected, thereby avoiding global desynchronization. This approach achieves on-demand local resynchronization for the first time, preserving all unambiguous tokens exactly as generated while simultaneously ensuring high security, capacity, and efficiency. Notably, it enables dual-channel steganographic communication with 100% reliability. Experimental results demonstrate extraction accuracy exceeding 99.7% in both Chinese and English settings, with negligible impact on text quality, embedding capacity, and computational efficiency compared to the original baseline, zero KL divergence indicating perfect distributional security, and end-to-end lossless recovery via the dual-channel mechanism.

Abstract

Generative linguistic steganography (GLS) enables covert communication by embedding secret messages into the natural language generation process. In practical deployment, however, GLS is vulnerable to tokenization ambiguity: the same surface text may be re-tokenized into a different token sequence at the receiver, breaking the shared decoding state between the communicating parties so that a single local mismatch can propagate into complete extraction failure. Existing solutions either remove ambiguous tokens -- distorting the generation distribution and compromising security -- or preserve the distribution at the cost of substantially reduced embedding capacity or prohibitive runtime overhead. To address this issue, we propose ReTokSync (Re-Tokenization Synchronization), a self-synchronizing disambiguation framework that monitors the receiver-view tokenization during generation and triggers a corrective reset only when ambiguity actually occurs. By confining the effect of tokenization ambiguity to sparse residual bit errors rather than global desynchronization, ReTokSync leaves ambiguity-free positions entirely untouched and remains compatible with the underlying steganographic algorithm. Experiments on both English and Chinese settings show that ReTokSync stays closest to the steganographic baseline in distributional security (zero KL divergence), text quality, embedding capacity, and runtime, while achieving extraction accuracy above 99.7\%. Building on this property, we further develop a two-channel covert communication mechanism in which ReTokSync serves as the primary channel and a reliable auxiliary channel corrects the remaining errors, achieving 100\% end-to-end recovery across all evaluated configurations.

Problem

• tokenization ambiguity
• generative linguistic steganography
• covert communication
• decoding desynchronization

Innovation

• ReTokSync
• tokenization disambiguation
• generative linguistic steganography
• self-synchronization
• covert communication