arXiv subjects

Minhao Bai

Publications and source records attributed to Minhao Bai.

9 recordsLinked to original sources

arXiv2026

Provably Secure Steganography Based on List Decoding

Steganography embeds secret messages in seemingly innocuous carriers for covert communication under surveillance. Current Provably Secure Steganography (PSS) schemes based on language models can guarantee computational indistinguishability between the covertext and stegotext. However, achieving high embedding capacity remains a challenge for existing PSS. The inefficient entropy utilization renders them not well-suited for Large Language Models (LLMs), whose inherent low-entropy tendencies severely constrain feasible embedding capacity. To address this, we propose a provably secure steganography scheme with a theoretically proved high capacity. Our scheme is based on the concept of list decoding: it maintains a set of candidates that contain the correct secret message, instead of directly finding the correct message with more effort. This strategy fully utilizes the information content of the generated text, yielding higher capacity. To ensure the correctness of our scheme, we further introduce a suffix-matching mechanism to distinguish the correct secret message from the candidates. We provide theoretical proofs for both the security and correctness of our scheme, alongside a derivation of its theoretical capacity lower bound. Our approach is plug-and-play, requiring only a direct replacement of the model's standard random sampling module. Experiments on three LLMs and seven PSS baselines demonstrate that our method achieves computational efficiency comparable to prior PSS schemes while delivering a substantial improvement in embedding capacity.

cs.CR↗

arXiv2025

Provable Secure Steganography Based on Adaptive Dynamic Sampling

The security of private communication is increasingly at risk due to widespread surveillance. Steganography, a technique for embedding secret messages within innocuous carriers, enables covert communication over monitored channels. Provably Secure Steganography (PSS), which ensures computational indistinguishability between the normal model output and steganography output, is the state-of-the-art in this field. However, current PSS methods often require obtaining the explicit distributions of the model. In this paper, we propose a provably secure steganography scheme that only requires a model API that accepts a seed as input. Our core mechanism involves sampling a candidate set of tokens and constructing a map from possible message bit strings to these tokens. The output token is selected by applying this mapping to the real secret message, which provably preserves the original model's distribution. To ensure correct decoding, we address collision cases, where multiple candidate messages map to the same token, by maintaining and strategically expanding a dynamic collision set within a bounded size range. Extensive evaluations of three real-world datasets and three large language models demonstrate that our sampling-based method is comparable with existing PSS methods in efficiency and capacity.

cs.CR↗

arXiv2025

Shifting-Merging: Secure, High-Capacity and Efficient Steganography via Large Language Models

In the face of escalating surveillance and censorship within the cyberspace, the sanctity of personal privacy has come under siege, necessitating the development of steganography, which offers a way to securely hide messages within innocent-looking texts. Previous methods alternate the texts to hide private massages, which is not secure. Large Language Models (LLMs) provide high-quality and explicit distribution, which is an available mathematical tool for secure steganography methods. However, existing attempts fail to achieve high capacity, time efficiency and correctness simultaneously, and their strongly coupling designs leave little room for refining them to achieve better performance. To provide a secure, high-capacity and efficient steganography method, we introduce ShiMer. Specifically, ShiMer pseudorandomly shifts the probability interval of the LLM's distribution to obtain a private distribution, and samples a token according to the private bits. ShiMer produced steganographic texts are indistinguishable in quality from the normal texts directly generated by the language model. To further enhance the capacity of ShiMer, we design a reordering algorithm to minimize the occurrence of interval splitting during decoding phase. Experimental results indicate that our method achieves the highest capacity and efficiency among existing secure steganography techniques.

cs.CR↗

arXiv2024

Let Watermarks Speak: A Robust and Unforgeable Watermark for Language Models

Watermarking is an effective way to trace model-generated content. Current watermark methods cannot resist forgery attacks, such as a deceptive claim that the model-generated content is a response to a fabricated prompt. None of them can be made unforgeable without degrading robustness. Unforgeability demands that the watermarked output is not only detectable but also verifiable for integrity, indicating whether it has been modified. This underscores the necessity and significance of a multi-bit watermarking scheme. Recent works try to build multi-bit scheme based on existing zero-bit watermarking scheme, but they either degrades the robustness or brings a significant computational burden. We aim to design a novel single-bit watermark scheme, which provides the ability to embed 2 different watermark signals. This paper's main contribution is that we are the first to propose an undetectable, robust, single-bit watermarking scheme. It has a comparable robustness to the most advanced zero-bit watermarking schemes. Then we construct a multi-bit watermarking scheme to use the hash value of prompt or the newest generated content as the watermark signals, and embed them into the following content, which guarantees the unforgeability. Additionally, we provide sufficient experiments on some popular language models, while the other advanced methods with provable guarantees do not often provide. The results show that our method is practically effective and robust.

cs.CR↗

arXiv2024

Semantic Steganography: A Framework for Robust and High-Capacity Information Hiding using Large Language Models

In the era of Large Language Models (LLMs), generative linguistic steganography has become a prevalent technique for hiding information within model-generated texts. However, traditional steganography methods struggle to effectively align steganographic texts with original model-generated texts due to the lower entropy of the predicted probability distribution of LLMs. This results in a decrease in embedding capacity and poses challenges for decoding stegos in real-world communication channels. To address these challenges, we propose a semantic steganography framework based on LLMs, which construct a semantic space and map secret messages onto this space using ontology-entity trees. This framework offers robustness and reliability for transmission in complex channels, as well as resistance to text rendering and word blocking. Additionally, the stegos generated by our framework are indistinguishable from the covers and achieve a higher embedding capacity compared to state-of-the-art steganography methods, while producing higher quality stegos.

cs.CR↗

arXiv2024

DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models

Discrete diffusion models have achieved success in tasks like image generation and masked language modeling but face limitations in controlled content editing. We introduce DICE (Discrete Inversion for Controllable Editing), the first approach to enable precise inversion for discrete diffusion models, including multinomial diffusion and masked generative models. By recording noise sequences and masking patterns during the reverse diffusion process, DICE enables accurate reconstruction and flexible editing of discrete data without the need for predefined masks or attention manipulation. We demonstrate the effectiveness of DICE across both image and text domains, evaluating it on models such as VQ-Diffusion, Paella, and RoBERTa. Our results show that DICE preserves high data fidelity while enhancing editing capabilities, offering new opportunities for fine-grained content manipulation in discrete spaces.

cs.CV↗

arXiv2024

Provably Robust and Secure Steganography in Asymmetric Resource Scenario

To circumvent the unbridled and ever-encroaching surveillance and censorship in cyberspace, steganography has garnered attention for its ability to hide private information in innocent-looking carriers. Current provably secure steganography approaches require a pair of encoder and decoder to hide and extract private messages, both of which must run the same model with the same input to obtain identical distributions. These requirements pose significant challenges to the practical implementation of steganography, including limited access to powerful hardware and the intolerance of any changes to the shared input. To relax the limitation of hardware and solve the challenge of vulnerable shared input, a novel and practically significant scenario with asymmetric resource should be considered, where only the encoder is high-resource and accessible to powerful models while the decoder can only read the steganographic carriers without any other model's input. This paper proposes a novel provably robust and secure steganography framework for the asymmetric resource setting. Specifically, the encoder uses various permutations of distribution to hide secret bits, while the decoder relies on a sampling function to extract the hidden bits by guessing the permutation used. Further, the sampling function only takes the steganographic carrier as input, which makes the decoder independent of model's input and model itself. A comprehensive assessment of applying our framework to generative models substantiates its effectiveness. Our implementation demonstrates robustness when transmitting over binary symmetric channels with errors.

cs.CR↗

arXiv2024

ModelShield: Adaptive and Robust Watermark against Model Extraction Attack

Large language models (LLMs) demonstrate general intelligence across a variety of machine learning tasks, thereby enhancing the commercial value of their intellectual property (IP). To protect this IP, model owners typically allow user access only in a black-box manner, however, adversaries can still utilize model extraction attacks to steal the model intelligence encoded in model generation. Watermarking technology offers a promising solution for defending against such attacks by embedding unique identifiers into the model-generated content. However, existing watermarking methods often compromise the quality of generated content due to heuristic alterations and lack robust mechanisms to counteract adversarial strategies, thus limiting their practicality in real-world scenarios. In this paper, we introduce an adaptive and robust watermarking method (named ModelShield) to protect the IP of LLMs. Our method incorporates a self-watermarking mechanism that allows LLMs to autonomously insert watermarks into their generated content to avoid the degradation of model content. We also propose a robust watermark detection mechanism capable of effectively identifying watermark signals under the interference of varying adversarial strategies. Besides, ModelShield is a plug-and-play method that does not require additional model training, enhancing its applicability in LLM deployments. Extensive evaluations on two real-world datasets and three LLMs demonstrate that our method surpasses existing methods in terms of defense effectiveness and robustness while significantly reducing the degradation of watermarking on the model-generated content.

cs.CR↗

arXiv2024

Learnable Linguistic Watermarks for Tracing Model Extraction Attacks on Large Language Models

In the rapidly evolving domain of artificial intelligence, safeguarding the intellectual property of Large Language Models (LLMs) is increasingly crucial. Current watermarking techniques against model extraction attacks, which rely on signal insertion in model logits or post-processing of generated text, remain largely heuristic. We propose a novel method for embedding learnable linguistic watermarks in LLMs, aimed at tracing and preventing model extraction attacks. Our approach subtly modifies the LLM's output distribution by introducing controlled noise into token frequency distributions, embedding an statistically identifiable controllable watermark.We leverage statistical hypothesis testing and information theory, particularly focusing on Kullback-Leibler Divergence, to differentiate between original and modified distributions effectively. Our watermarking method strikes a delicate well balance between robustness and output quality, maintaining low false positive/negative rates and preserving the LLM's original performance.

cs.CR↗