HiTMS: A High-Throughput Multi-Stream Linguistic Steganography Framework

2026-07-26

AI Summary

This work addresses limitations in existing generative linguistic steganography methods, which typically employ single-stream designs that hinder batched multi-stream inference and fail to conceal slot occupancy and payload status. To overcome these issues, the authors propose a multi-stream steganographic framework supporting multi-turn interactions. By leveraging a self-describing frame structure and a key-derived slot scheduling mechanism, the framework enables parallel embedding and extraction of multiple message streams within a single batched model invocation, substantially amortizing computational overhead. Integrated with a decoy-padding strategy, the method accurately recovers secret messages while concealing the number of active streams, and remains compatible with arbitrary language models and steganographic encoders. Experiments demonstrate up to a 4.3× speedup in embedding and extraction under an 8-stream configuration, reducing steganalysis AUROC from 0.681 to 0.601, with sustained throughput advantages even at 64 concurrent streams.

Abstract

Generative linguistic steganography conceals secret bits within the sampling randomness of large language models. Existing schemes are single-stream, conveying an entire secret through a single response to a single prompt. This convention incurs two limitations: it provides no protocol-level support for batched multi-stream inference, and naive co-batching does not conceal slot occupancy or payload completion. We propose HiTMS, which distributes a secret across multiple responses produced jointly over successive rounds of interaction. Each round embeds and extracts several streams within a single batched call, thereby amortizing the cost of model invocation and substantially improving throughput. To ensure recoverability, HiTMS wraps each response in a self-describing frame and employs a key-derived schedule that binds streams to slots and fills unused slots with decoys, guaranteeing exact recovery while concealing the number of active streams. The framework is agnostic to both the language model and the steganographic coder. Across eight dataset-model-coder settings, eight-stream HiTMS achieves up to 4.3 times higher embedding and extraction speeds than single-stream baselines, while reducing the steganalyzer AUROC from 0.681 to 0.601 on average. Additional experiments with 4 to 64 streams demonstrate sustained throughput gains as concurrency increases. GitHub repository for this work is https://github.com/ryehr/HiTMS_steganography.

Problem

• linguistic steganography
• multi-stream
• batched inference
• slot occupancy
• payload completion

Innovation

• multi-stream steganography
• high-throughput
• batched inference
• self-describing frame
• key-derived scheduling