arXiv subjects

Hanzhou Wu

Publications and source records attributed to Hanzhou Wu.

At least 19 recordsLinked to original sources

arXiv2026

Beyond QA Matching: Perturbation-Response Fingerprinting via Probability Distributions for Large Language Models

Large language models are often instruction-tuned, specialized, quantized, or otherwise transformed, making fine-grained provenance difficult. In this paper, we introduce BReF, a training-free fingerprint that compares how probability distributions over four answer-option labels A/B/C/D move under controlled textual perturbations. For each pair of models, BReF selects 25 jointly responsive probes and compares their perturbation log-ratio (PLR) response directions by global cosine similarity. On a unified benchmark with 34 checkpoints, 22 documented direct-parent relations, and 411 suspect-candidate pairs, BReF retrieves the documented parent in 22/22 cases (MRR=1.0000), with DP-DF AUC 1.0000. Same-family discrimination is harder (DP-SF AUC 0.8969), and paired tests show a significant exact-retrieval gain over a magnitude-only Top-25 control. Together with static, random-probe, permutation, calibration, and transformation-level controls, the results show that strong pooled separation does not guarantee correct parent ranking among closely related checkpoints, verifying the superiority of our work.

cs.CR↗

arXiv2026

Code-Level Cost Function Generation for Spatial Image Steganography Using RAG-Enhanced Large Language Models

Designing cost functions of adaptive steganography traditionally requires extensive manual tuning, while deep learning methods lack interpretability. Although large language models (LLMs) offer an automated alternative via evolutionary generation, they often violate domain specific mathematical constraints due to a lack of explicit domain knowledge. To address this problem, we propose a novel evolutionary system focused on exploiting Retrieval-Augmented Generation (RAG) enhanced LLMs for the automatic code-level generation of spatial steganography cost functions. This system incorporates a core Self Evolving RAG (SE-RAG) module, wherein a Code Semantic Signature (CSS) translates procedural code into aligned queries, retrieving explicit guidance from static literature and dynamic experience knowledge bases to steer the LLM generation process. A dedicated feedback mechanism then continuously refines the dynamic knowledge base with successful optimization strategies. Extensive experiments on the BOSSBase and BOWS2 datasets demonstrate that the proposed framework consistently achieves higher steganographic security than existing automatically designed methods, and increases the average code execution rate by 46.3% while reducing the search cost by 26.1%, thereby highlighting the effectiveness, efficiency, and potential of combining LLMs with domain-specific knowledge in the field of automatic steganographic algorithm generation.

cs.CR↗

arXiv2026

R-CoT: A Reasoning-Layer Watermark via Redundant Chain-of-Thought in Large Language Models

Large language models (LLMs) are widely deployed in multiple scenarios due to reasoning capabilities. In order to prevent the models from being misused, watermarking is generally employed to ensure ownership. However, most existing watermarking methods rely on superficial modifications to the model's output distribution, rendering the watermark vulnerable to perturbation and removal. To overcome this challenge, this paper introduces a reasoning-layer framework termed Redundant Chain-of-Thought (R-CoT), which embeds watermarks into the reasoning path. A dual-trajectory optimization mechanism based on GRPO enables the native and the watermark reasoning path to coexist within a shared parameter space, internalizing the watermark as a distinct reasoning policy. Therefore, the watermark is embedded into the model's stable reasoning path, avoiding the watermark failure caused by output-level perturbations. Experimental results show that, compared with existing methods, R-CoT achieves high watermark effectiveness and strong robustness. Under fine-tuning and other post-training operations, the true positive rate (TPR) consistently remains above 95%, exhibiting only marginal degradation.

cs.CR↗

arXiv2025

Defining Cost Function of Steganography with Large Language Models

In this paper, we make the first attempt towards defining cost function of steganography with large language models (LLMs), which is totally different from previous works that rely heavily on expert knowledge or require large-scale datasets for cost learning. To achieve this goal, a two-stage strategy combining LLM-guided program synthesis with evolutionary search is applied in the proposed method. In the first stage, a certain number of cost functions in the form of computer programs are synthesized from LLM responses to structured prompts. These cost functions are then evaluated with pretrained steganalysis models so that candidate cost functions suited to steganography can be collected. In the second stage, by retraining a steganalysis model for each candidate cost function, the optimal cost function(s) can be determined according to the detection accuracy. This two-stage strategy is performed by an iterative fashion so that the best cost function can be collected at the last iteration. Experiments show that the proposed method enables LLMs to design new cost functions of steganography that significantly outperform existing works in terms of resisting steganalysis tools, which verifies the superiority of the proposed method. To the best knowledge of the authors, this is the first work applying LLMs to the design of advanced cost function of steganography, which presents a novel perspective for steganography design and may shed light on further research.

cs.CR↗

arXiv2025

Rotation, Scale, and Translation Resilient Black-box Fingerprinting for Intellectual Property Protection of EaaS Models

Feature embedding has become a cornerstone technology for processing high-dimensional and complex data, which results in that Embedding as a Service (EaaS) models have been widely deployed in the cloud. To protect the intellectual property of EaaS models, existing methods apply digital watermarking to inject specific backdoor triggers into EaaS models by modifying training samples or network parameters. However, these methods inevitably produce detectable patterns through semantic analysis and exhibit susceptibility to geometric transformations including rotation, scaling, and translation (RST). To address this problem, we propose a fingerprinting framework for EaaS models, rather than merely refining existing watermarking techniques. Different from watermarking techniques, the proposed method establishes EaaS model ownership through geometric analysis of embedding space's topological structure, rather than relying on the modified training samples or triggers. The key innovation lies in modeling the victim and suspicious embeddings as point clouds, allowing us to perform robust spatial alignment and similarity measurement, which inherently resists RST attacks. Experimental results evaluated on visual and textual embedding tasks verify the superiority and applicability. This research reveals inherent characteristics of EaaS models and provides a promising solution for ownership verification of EaaS models under the black-box scenario.

cs.CR↗

arXiv2025

An Information Asymmetry Game for Trigger-based DNN Model Watermarking

As a valuable digital product, deep neural networks (DNNs) face increasingly severe threats to the intellectual property, making it necessary to develop effective technical measures to protect them. Trigger-based watermarking methods achieve copyright protection by embedding triggers into the host DNNs. However, the attacker may remove the watermark by pruning or fine-tuning. We model this interaction as a game under conditions of information asymmetry, namely, the defender embeds a secret watermark with private knowledge, while the attacker can only access the watermarked model and seek removal. We define strategies, costs, and utilities for both players, derive the attacker's optimal pruning budget, and establish an exponential lower bound on the accuracy of watermark detection after attack. Experimental results demonstrate the feasibility of the watermarked model, and indicate that sparse watermarking can resist removal with negligible accuracy loss. This study highlights the effectiveness of game-theoretic analysis in guiding the design of robust watermarking schemes for model copyright protection.

cs.CR↗

arXiv2025

Trigger Where It Hurts: Unveiling Hidden Backdoors through Sensitivity with Sensitron

Backdoor attacks pose a significant security threat to natural language processing (NLP) systems, but existing methods lack explainable trigger mechanisms and fail to quantitatively model vulnerability patterns. This work pioneers the quantitative connection between explainable artificial intelligence (XAI) and backdoor attacks, introducing Sensitron, a novel modular framework for crafting stealthy and robust backdoor triggers. Sensitron employs a progressive refinement approach where Dynamic Meta-Sensitivity Analysis (DMSA) first identifies potentially vulnerable input tokens, Hierarchical SHAP Estimation (H-SHAP) then provides explainable attribution to precisely pinpoint the most influential tokens, and finally a Plug-and-Rank mechanism that generates contextually appropriate triggers. We establish the first mathematical correlation (Sensitivity Ranking Correlation, SRC=0.83) between explainability scores and empirical attack success, enabling precise targeting of model vulnerabilities. Sensitron achieves 97.8% Attack Success Rate (ASR) (+5.8% over state-of-the-art (SOTA)) with 85.4% ASR at 0.1% poisoning rate, demonstrating robust resistance against multiple SOTA defenses. This work reveals fundamental NLP vulnerabilities and provides new attack vectors through weaponized explainability.

cs.CR↗

arXiv2025

Yet Another Watermark for Large Language Models

Existing watermarking methods for large language models (LLMs) mainly embed watermark by adjusting the token sampling prediction or post-processing, lacking intrinsic coupling with LLMs, which may significantly reduce the semantic quality of the generated marked texts. Traditional watermarking methods based on training or fine-tuning may be extendable to LLMs. However, most of them are limited to the white-box scenario, or very time-consuming due to the massive parameters of LLMs. In this paper, we present a new watermarking framework for LLMs, where the watermark is embedded into the LLM by manipulating the internal parameters of the LLM, and can be extracted from the generated text without accessing the LLM. Comparing with related methods, the proposed method entangles the watermark with the intrinsic parameters of the LLM, which better balances the robustness and imperceptibility of the watermark. Moreover, the proposed method enables us to extract the watermark under the black-box scenario, which is computationally efficient for use. Experimental results have also verified the feasibility, superiority and practicality. This work provides a new perspective different from mainstream works, which may shed light on future research.

cs.CR↗

arXiv2025

Watermarking Quantum Neural Networks Based on Sample Grouped and Paired Training

Quantum neural networks (QNNs) leverage quantum computing to create powerful and efficient artificial intelligence models capable of solving complex problems significantly faster than traditional computers. With the fast development of quantum hardware technology, such as superconducting qubits, trapped ions, and integrated photonics, quantum computers may become reality, accelerating the applications of QNNs. However, preparing quantum circuits and optimizing parameters for QNNs require quantum hardware support, expertise, and high-quality data. How to protect intellectual property (IP) of QNNs becomes an urgent problem to be solved in the era of quantum computing. We make the first attempt towards IP protection of QNNs by watermarking. To this purpose, we collect classical clean samples and trigger ones, each of which is generated by adding a perturbation to a clean sample, associated with a label different from the ground-truth one. The host QNN, consisting of quantum encoding, quantum state transformation, and quantum measurement, is then trained from scratch with the clean samples and trigger ones, resulting in a watermarked QNN model. During training, we introduce sample grouped and paired training to ensure that the performance on the downstream task can be maintained while achieving good performance for watermark extraction. When disputes arise, by collecting a mini-set of trigger samples, the hidden watermark can be extracted by analyzing the prediction results of the target model corresponding to the trigger samples, without accessing the internal details of the target QNN model, thereby verifying the ownership of the model. Experiments have verified the superiority and applicability of this work.

cs.CR↗

arXiv2025

ShadowCoT: Cognitive Hijacking for Stealthy Reasoning Backdoors in LLMs

Chain-of-Thought (CoT) enhances an LLM's ability to perform complex reasoning tasks, but it also introduces new security issues. In this work, we present ShadowCoT, a novel backdoor attack framework that targets the internal reasoning mechanism of LLMs. Unlike prior token-level or prompt-based attacks, ShadowCoT directly manipulates the model's cognitive reasoning path, enabling it to hijack multi-step reasoning chains and produce logically coherent but adversarial outcomes. By conditioning on internal reasoning states, ShadowCoT learns to recognize and selectively disrupt key reasoning steps, effectively mounting a self-reflective cognitive attack within the target model. Our approach introduces a lightweight yet effective multi-stage injection pipeline, which selectively rewires attention pathways and perturbs intermediate representations with minimal parameter overhead (only 0.15% updated). ShadowCoT further leverages reinforcement learning and reasoning chain pollution (RCP) to autonomously synthesize stealthy adversarial CoTs that remain undetectable to advanced defenses. Extensive experiments across diverse reasoning benchmarks and LLMs show that ShadowCoT consistently achieves high Attack Success Rate (94.4%) and Hijacking Success Rate (88.4%) while preserving benign performance. These results reveal an emergent class of cognition-level threats and highlight the urgent need for defenses beyond shallow surface-level consistency.

cs.CR↗

arXiv2025

A Game Between the Defender and the Attacker for Trigger-based Black-box Model Watermarking

Watermarking deep neural network (DNN) models has attracted a great deal of attention and interest in recent years because of the increasing demand to protect the intellectual property of DNN models. Many practical algorithms have been proposed by covertly embedding a secret watermark into a given DNN model through either parametric/structural modulation or backdooring against intellectual property infringement from the attacker while preserving the model performance on the original task. Despite the performance of these approaches, the lack of basic research restricts the algorithmic design to either a trial-based method or a data-driven technique. This has motivated the authors in this paper to introduce a game between the model attacker and the model defender for trigger-based black-box model watermarking. For each of the two players, we construct the payoff function and determine the optimal response, which enriches the theoretical foundation of model watermarking and may inspire us to develop novel schemes in the future.

cs.CR↗

arXiv2024

Discriminative Anchor Learning for Efficient Multi-view Clustering

Multi-view clustering aims to study the complementary information across views and discover the underlying structure. For solving the relatively high computational cost for the existing approaches, works based on anchor have been presented recently. Even with acceptable clustering performance, these methods tend to map the original representation from multiple views into a fixed shared graph based on the original dataset. However, most studies ignore the discriminative property of the learned anchors, which ruin the representation capability of the built model. Moreover, the complementary information among anchors across views is neglected to be ensured by simply learning the shared anchor graph without considering the quality of view-specific anchors. In this paper, we propose discriminative anchor learning for multi-view clustering (DALMC) for handling the above issues. We learn discriminative view-specific feature representations according to the original dataset and build anchors from different views based on these representations, which increase the quality of the shared anchor graph. The discriminative feature learning and consensus anchor graph construction are integrated into a unified framework to improve each other for realizing the refinement. The optimal anchors from multiple views and the consensus anchor graph are learned with the orthogonal constraints. We give an iterative algorithm to deal with the formulated problem. Extensive experiments on different datasets show the effectiveness and efficiency of our method compared with other methods.

cs.LG↗

arXiv2024

TCNFormer: Temporal Convolutional Network Former for Short-Term Wind Speed Forecasting

Global environmental challenges and rising energy demands have led to extensive exploration of wind energy technologies. Accurate wind speed forecasting (WSF) is crucial for optimizing wind energy capture and ensuring system stability. However, predicting wind speed remains challenging due to its inherent randomness, fluctuation, and unpredictability. This study proposes the Temporal Convolutional Network Former (TCNFormer) for short-term (12-hour) wind speed forecasting. The TCNFormer integrates the Temporal Convolutional Network (TCN) and transformer encoder to capture the spatio-temporal features of wind speed. The transformer encoder consists of two distinct attention mechanisms: causal temporal multi-head self-attention (CT-MSA) and temporal external attention (TEA). CT-MSA ensures that the output of a step derives only from previous steps, i.e., causality. Locality is also introduced to improve efficiency. TEA explores potential relationships between different sample sequences in wind speed data. This study utilizes wind speed data from the NASA Prediction of Worldwide Energy Resources (NASA POWER) of Patenga Sea Beach, Chittagong, Bangladesh (latitude 22.2352{\deg} N, longitude 91.7914{\deg} E) over a year (six seasons). The findings indicate that the TCNFormer outperforms state-of-the-art models in prediction accuracy. The proposed TCNFormer presents a promising method for spatio-temporal WSF and may achieve desirable performance in real-world applications of wind power systems.

cs.LG↗

arXiv2024

A Fingerprint for Large Language Models

Recent advances confirm that large language models (LLMs) can achieve state-of-the-art performance across various tasks. However, due to the resource-intensive nature of training LLMs from scratch, it is urgent and crucial to protect the intellectual property of LLMs against infringement. This has motivated the authors in this paper to propose a novel black-box fingerprinting technique for LLMs. We firstly demonstrate that the outputs of LLMs span a unique vector space associated with each model. We model the problem of fingerprint authentication as the task of evaluating the similarity between the space of the victim model and the space of the suspect model. To tackle with this problem, we introduce two solutions: the first determines whether suspect outputs lie within the victim's subspace, enabling fast infringement detection; the second reconstructs a joint subspace to detect models modified via parameter-efficient fine-tuning (PEFT). Experiments indicate that the proposed method achieves superior performance in fingerprint verification and robustness against the PEFT attacks. This work reveals inherent characteristics of LLMs and provides a promising solution for protecting LLMs, ensuring efficiency, generality and practicality.

cs.CR↗

arXiv2024

Transferable Watermarking to Self-supervised Pre-trained Graph Encoders by Trigger Embeddings

Recent years have witnessed the prosperous development of Graph Self-supervised Learning (GSSL), which enables to pre-train transferable foundation graph encoders. However, the easy-to-plug-in nature of such encoders makes them vulnerable to copyright infringement. To address this issue, we develop a novel watermarking framework to protect graph encoders in GSSL settings. The key idea is to force the encoder to map a set of specially crafted trigger instances into a unique compact cluster in the outputted embedding space during model pre-training. Consequently, when the encoder is stolen and concatenated with any downstream classifiers, the resulting model inherits the `backdoor' of the encoder and predicts the trigger instances to be in a single category with high probability regardless of the ground truth. Experimental results have shown that, the embedded watermark can be transferred to various downstream tasks in black-box settings, including node classification, link prediction and community detection, which forms a reliable watermark verification system for GSSL in reality. This approach also shows satisfactory performance in terms of model fidelity, reliability and robustness.

cs.CR↗

arXiv2023

StegGuard: Fingerprinting Self-supervised Pre-trained Encoders via Secrets Embeder and Extractor

In this work, we propose StegGuard, a novel fingerprinting mechanism to verify the ownership of the suspect pre-trained encoder using steganography. A critical perspective in StegGuard is that the unique characteristic of the transformation from an image to an embedding, conducted by the pre-trained encoder, can be equivalently exposed how an embeder embeds secrets into images and how an extractor extracts the secrets from encoder's embeddings with a tolerable error after the secrets are subjected to the encoder's transformation. While each independent encoder has a distinct transformation, the piracy encoder has a similar transformation to the victim. Based on these, we learn a pair of secrets embeder and extractor as the fingerprint for the victim encoder. We introduce a frequency-domain channel attention embedding block into the embeder to adaptively embed secrets into suitable frequency bands. During verification, if the secrets embedded into the query images can be extracted with an acceptable error from the suspect encoder's embeddings, the suspect encoder is determined as piracy, otherwise independent. Extensive experiments demonstrate that depending on a very limited number of query images, StegGuard can reliably identify across varied independent encoders, and is robust against model stealing related attacks including model extraction, fine-tuning, pruning, embedding noising and shuffle.

cs.CR↗

arXiv2023

Generative Model Watermarking Suppressing High-Frequency Artifacts

Protecting deep neural networks (DNNs) against intellectual property (IP) infringement has attracted an increasing attention in recent years. Recent advances focus on IP protection of generative models, which embed the watermark information into the image generated by the model to be protected. Although the generated marked image has good visual quality, it introduces noticeable artifacts to the marked image in high-frequency area, which severely impairs the imperceptibility of the watermark and thereby reduces the security of the watermarking system. To deal with this problem, in this paper, we propose a novel framework for generative model watermarking that can suppress those high-frequency artifacts. The main idea of the proposed framework is to design a new watermark embedding network that can suppress high-frequency artifacts by applying anti-aliasing. To realize anti-aliasing, we use low-pass filtering for the internal sampling layers of the new watermark embedding network. Meanwhile, joint loss optimization and adversarial training are applied to enhance the effectiveness and robustness. Experimental results indicate that the marked model not only maintains the performance very well on the original task, but also demonstrates better imperceptibility and robustness on the watermarking task. This work reveals the importance of suppressing high-frequency artifacts for enhancing imperceptibility and security of generative model watermarking.

cs.CR↗

arXiv2023

JPEG Steganalysis Based on Steganographic Feature Enhancement and Graph Attention Learning

The purpose of image steganalysis is to determine whether the carrier image contains hidden information or not. Since JEPG is the most commonly used image format over social networks, steganalysis in JPEG images is also the most urgently needed to be explored. However, in order to detect whether secret information is hidden within JEPG images, the majority of existing algorithms are designed in conjunction with the popular computer vision related networks, without considering the key characteristics appeared in image steganalysis. It is crucial that the steganographic signal, as an extremely weak signal, can be enhanced during its representation learning process. Motivated by this insight, in this paper, we introduce a novel representation learning algorithm for JPEG steganalysis that is mainly consisting of a graph attention learning module and a feature enhancement module. The graph attention learning module is designed to avoid global feature loss caused by the local feature learning of convolutional neural network and reliance on depth stacking to extend the perceptual domain. The feature enhancement module is applied to prevent the stacking of convolutional layers from weakening the steganographic information. In addition, pretraining as a way to initialize the network weights with a large-scale dataset is utilized to enhance the ability of the network to extract discriminative features. We advocate pretraining with ALASKA2 for the model trained with BOSSBase+BOWS2. The experimental results indicate that the proposed algorithm outperforms previous arts in terms of detection accuracy, which has verified the superiority and applicability of the proposed work.

cs.MM↗