LLMs have blurred the lines between automated and human text generation, an advancement that is advantageous for text steganography. Current methods, however, are not fit for LLMs due to most users’ lack of access to the underlying models. This paper proposes LLM-Stega, a black-box generative text steganography method, enabling covert communication through LLM interfaces.

• Introduces LLM-Stega for secure covert communication using LLM interfaces.
• Constructs a keyword set and designs encrypted steganographic mapping for secret message embedding.
• Proposes an optimization mechanism based on reject sampling to ensure accurate message extraction and semantic richness.
• Demonstrates superior performance to current state-of-the-art methods.

This research showcases the potential of using LLMs for secure communication, setting the stage for new applications in fields requiring privacy and discretion. The approach can strengthen data protection and open up avenues for innovative cybersecurity solutions. Discover the full methodology.