Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generative Text Steganography with Large Language Model

Jiaxuan Wu, Zhengxian Wu, Yiming Xue, Juan Wen, Wanli Peng

TL;DR

A black-box generative text steganographic method based on the user interfaces of large language models, which is called LLM-Stega, to ensure secure covert communication between Alice and Bob by using the user interfaces of LLMs.

Abstract

Recent advances in large language models (LLMs) have blurred the boundary of high-quality text generation between humans and machines, which is favorable for generative text steganography. While, current advanced steganographic mapping is not suitable for LLMs since most users are restricted to accessing only the black-box API or user interface of the LLMs, thereby lacking access to the training vocabulary and its sampling probabilities. In this paper, we explore a black-box generative text steganographic method based on the user interfaces of large language models, which is called LLM-Stega. The main goal of LLM-Stega is that the secure covert communication between Alice (sender) and Bob (receiver) is conducted by using the user interfaces of LLMs. Specifically, We first construct a keyword set and design a new encrypted steganographic mapping to embed secret messages. Furthermore, to guarantee accurate extraction of secret messages and rich semantics of generated stego texts, an optimization mechanism based on reject sampling is proposed. Comprehensive experiments demonstrate that the proposed LLM-Stega outperforms current state-of-the-art methods.

Generative Text Steganography with Large Language Model

TL;DR

A black-box generative text steganographic method based on the user interfaces of large language models, which is called LLM-Stega, to ensure secure covert communication between Alice and Bob by using the user interfaces of LLMs.

Abstract

Recent advances in large language models (LLMs) have blurred the boundary of high-quality text generation between humans and machines, which is favorable for generative text steganography. While, current advanced steganographic mapping is not suitable for LLMs since most users are restricted to accessing only the black-box API or user interface of the LLMs, thereby lacking access to the training vocabulary and its sampling probabilities. In this paper, we explore a black-box generative text steganographic method based on the user interfaces of large language models, which is called LLM-Stega. The main goal of LLM-Stega is that the secure covert communication between Alice (sender) and Bob (receiver) is conducted by using the user interfaces of LLMs. Specifically, We first construct a keyword set and design a new encrypted steganographic mapping to embed secret messages. Furthermore, to guarantee accurate extraction of secret messages and rich semantics of generated stego texts, an optimization mechanism based on reject sampling is proposed. Comprehensive experiments demonstrate that the proposed LLM-Stega outperforms current state-of-the-art methods.
Paper Structure (14 sections, 5 equations, 4 figures, 4 tables, 1 algorithm)

This paper contains 14 sections, 5 equations, 4 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Generative text Steganography
  4. Large Language Models
  5. The Proposed Methodology
  6. Keyword Set Construction
  7. Encrypted Steganographic Mapping
  8. Steganographic Text Generation and Secret Message Extraction
  9. Experiment
  10. Experimental Setup
  11. Metrics
  12. Results and Analysis
  13. Ablation Study
  14. Conclusion

Figures (4)

  • Figure 1: (a) previous methods rely on a language model and steganographic mapping, which is a white-box method; (b) the proposed LLM-Stega generates stego texts by using large language models. The LLM-Stega directly uses the UIs of LLMs to embed and extract secret messages, which is a black-box method.
  • Figure 2: The overall framework of LLM-Stega. Lc-Idx and Re-Idx are the location index and repetition number of the selected keywords in the augmentation keyword set, respectively. Sec-Mes denotes the secret messages. The $w_i$ and $p_i$ denote the i-th word and i-th sampling probability in the keyword set, respectively.
  • Figure 3: The experimental results of the Anti-steganalysis ability
  • Figure 4: The experimental results of the Statistical Imperceptibility and the Human Evaluation.