Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks159 from public data
- Gaussian Shading: Provable Performance-Lossless Image Watermarking for Diffusion Models233
Gaussian Shading is proposed, a diffusion model watermarking technique that is both performance-lossless and training-free, while serving the dual purpose of copyright protection and tracing of offending content.
- 220
- LG-GAN: Label Guided Adversarial Network for Flexible Targeted Attack of Point Cloud Based Deep Networks130
The proposed LG-GAN can support flexible targeted attack on the fly while guaranteeing good attack performance and higher efficiency simultaneously, and is the first generation based 3D point cloud attack method.
- Adversarial Examples Against Deep Neural Network based Steganalysis108
A new strategy is proposed that constructs enhanced covers against neural networks with the technique of adversarial examples and makes a tradeoff between the two analysis systems to improve the comprehensive security.
- Watermarking Text Generated by Black-Box Language Models103
A watermarking framework for black-box language model usage scenarios and results under re-translation, polishing, word deletion, and synonym substitution attacks reveal that it is arduous to remove the watermark without compromising the original semantics.
- Tracing Text Provenance via Context-Aware Lexical Substitution102
A natural language watermarking scheme based on context-aware lexical substitution (LS) that employs BERT to suggest LS candidates by inferring the semantic relatedness between the candidates and the original sentence and has a better transferability than existing methods.
- Flow-Based Robust Watermarking with Invertible Noise Layer for Black-Box Distortions95
Compared with the state-of-the-art architecture, the visual quality of the proposed framework improves by 2dB and the extraction accuracy after JPEG compression improves by more than 4%.
- 92
- Toward Understanding and Boosting Adversarial Transferability From a Distribution Perspective89
This work proposes a novel method that crafts adversarial examples by manipulating the distribution of the image and can significantly improve the transferability of the crafted attacks and achieves state-of-the-art performance in both untargeted and targeted scenarios.
- 77
- Robust adaptive steganography based on generalized dither modulation and expanded embedding domain71
The purpose of this paper is to refine the robust steganographic scheme by considering asymmetric costs for different modification polarities and expanding the embedding domain for digital images, aiming to aggregate the modifications on the elements with small costs.
- A Closer Look at Machine Unlearning for Large Language Models63
To address the issue of inadequate evaluation of model outputs after unlearning, three additional metrics are introduced to evaluate token diversity, sentence semantics, and factual correctness and to alleviate issues in machine unlearning for LLMs.
- StegaDDPM: Generative Image Steganography based on Denoising Diffusion Probabilistic Model60
StegaDDPM outperforms state-of-the-art methods in terms of steganographic capacity, extraction accuracy, and security, and can securely conceal and accurately extract secret messages up to 9 bits per pixel.
- 58
- 57
- 55
- 53
- Provably Secure Generative Steganography Based on Autoregressive Model43
This paper proposes a novel framework for generative steganography based on autoregressive model, or rather, PixelCNN, and shows that the proposed method can resist the existing steganalytic methods.
- 39
- Adversarial batch image steganography against CNN-based pooled steganalysis37
This paper proposes a general framework of adversarial embedding for batch steganography, in which a new loss function is designed and the batch strategy is combined with adversarialembedding, and shows the proposed method significantly improves the security performance of batch Steganography against pooled steganalysis and keeps a high-security level against single image Steganalysis.
- Defining Joint Distortion for JPEG Steganography37
This paper inspects the embedding change from the spatial domain and proposes a principle of Block Boundary Continuity (BBC) for defining JPEG joint distortion, which aims to restrain blocking artifacts caused by inter-block adjacent modifications and thus effectively preserve the spatial continuity at block boundaries.
- 36
- LDStega: Practical and Robust Generative Image Steganography based on Latent Diffusion Models35
Experimental results demonstrate that LDStega achieves high extraction accuracy while controllably generating image content and saving the stego image in the widely used PNG and JPEG formats and outperforms state-of-the-art techniques in resisting common image attacks.
- 33
- 32
- DeNoL: A Few-Shot-Sample-Based Decoupling Noise Layer for Cross-channel Watermarking Robustness31
DeNoL is proposed, a decoupling noise layer for cross-channel simulation which only needs few-shot samples and can effectively simulate cross-Channel distortion with only 20 image pairs and assist in training a general and robust watermarking network.
- 31
- Text Fluoroscopy: Detecting LLM-Generated Text through Intrinsic Features29
Text Fluoroscopy is designed, a black-box method with better generalizability for detecting LLM-generated text by mining the intrinsic features of the text to be detected by mining the intrinsic features of the text to be detected.
- Three-Dimensional Mesh Steganography and Steganalysis: A Review28
A systematic survey of the literature on 3-D mesh steganography and steganalysis, and proposes a new taxonomy of steganographic algorithms with four categories: 1) two-state domain, 2) LSB domain, 3) permutation domain, and 4) transform domain.
- 27
- 27
- Controlling risks of AI in chemical science with agents26
SciGuard is proposed, an agent-based guardrail that employs large language models, tools and external knowledge to assess and control risks in scientific AI interactions, and achieves a state-of-the-art harmlessness score on red-teaming queries, while maintaining high performance on benign tasks, without sacrificing scientific knowledge.
- Self-Supervised Adversarial Training26
Experimental results show that the self- supervised representation outperforms its supervised version in respect of robustness and self-supervised adversarial training can further improve the defense ability efficiently.
- AutoPT: How Far Are We from the End2End Automated Web Penetration Testing?25
This work introduces AutoPT, an automated penetration testing agent based on the principle of PSM driven by LLMs, which utilizes the inherent inference ability of LLM and the constraint framework of state machines to address the limitations of penetration testing agents.
- InferDPT: Privacy-Preserving Inference for Closed-Box Large Language Models24
Experimental results across three datasets demonstrate that the text generation quality of <monospace>InferDPT</monospace> is comparable to that of non-private GPT-4, and RANTEXT surpasses existing state-of-the-art mechanisms, namely, SANTEXT+ and CUSTEXT+ in the trade-off between privacy and utility.
- DPIC: Decoupling Prompt and Intrinsic Characteristics for LLM Generated Text Detection24
The generation process is viewed as a coupled process of prompt and intrinsic characteristics of the generative model, and DPIC is proposed, which eliminates the prompt in the detection process, which allows the detector to focus on the intrinsic characteristics of the generative model.
- 24
- 23
- Silent Guardian: Protecting Text From Malicious Exploitation by Large Language Models22
Silent Guardian (SG) is introduced, a text protection mechanism against LLMs, which allows LLMs to refuse to generate responses when receiving protected text, preventing the malicious use of text from the source.
- 22
- 22
- Cover Reproducible Steganography via Deep Generative Models22
This work proposes cover-reproducible steganography where the source coding, e.g., arithmetic coding, serves as the steganographic code and the decoding process of arithmetic coding is used for message embedding and its encoding process is regarded as message extraction.
- Image Steganography and Style Transformation Based on Generative Adversarial Network21
The proposed scheme takes a content image, an art-style image, and messages to be embedded as inputs, processing them with an encoder–decoder model, and generates a styled image containing the secret messages at the same time.
- Imperceptible black-box waveform-level adversarial attack towards automatic speaker recognition21
This paper proposes a new type of adversarial examples by generating imperceptible adversarial samples for targeted attacks on black-box systems of automatic speaker recognition, Inspired by auditory masking, a regularization term adapting to the energy of speech waveform is proposed for generating imperceptive adversarial perturbations.
- SemGIR: Semantic-Guided Image Regeneration Based Method for AI-generated Image Detection and Attribution20
A semantic-guided image regeneration based method for AI-generated image detection, SemGIR first regenerates images through image-to-text followed by a text-to-image generation process, subsequently utilizing these re-generated image pairs to derive discriminative features.
- 19
- 18
- 17
- Provably Secure Disambiguating Neural Linguistic Steganography16
A novel secure disambiguation method named SyncPool is proposed, which effectively addresses the segmentation ambiguity problem and has the potential to significantly improve the reliability and security of neural linguistic steganography systems.
- 16
- Adversarial Examples Detection Beyond Image Space15
This work proposes a method beyond image space by a two-stream architecture, in which the image stream focuses on the pixel artifacts and the gradient stream copes with the confidence artifacts, which outperforms the existing methods under oblivious attacks and is verified effective to defend omniscient attacks.
- 15
- 14
- Robust Secret Image Sharing Resistant to JPEG Recompression Based on Stable Block Condition14
SBC-ROSIS (Robust Secret Image Sharing Scheme Resistant to JPEG Recompression Based on Stable Block Condition), a novel ROSIS scheme that utilizes a stable block condition to guarantee the invariance of discrete cosine transform (DCT) coefficients during JPEG recompression, significantly enhancing the robustness of the scheme.
- 14
- 13
- 13
- Invertible Image Dataset Protection12
An invertible Image Dataset Protection NET-work (IDP-Net) is developed to introduce slight and acceptable changes to the images within the dataset to better protect the security of image dataset against defensive methods compared to previ-ous methods.
- 11
- CodeWMBench: An Automated Benchmark for Code Watermarking Evaluation11
A comprehensive automated benchmark CodeWMBench is introduced for active detection of human-machine code through watermarking and the first assessment of these watermarking methods against code rewriting and retranslating attacks is conducted.
- 11
- Revisiting Entropy Regularization: Adaptive Coefficient Unlocks Its Potential for LLM Reinforcement Learning10
This work revisits entropy regularization in RLVR and argues that its potential has been largely underestimated, and proposes Adaptive Entropy Regularization (AER)--a framework that dynamically balances exploration and exploitation via three components: difficulty-aware coefficient allocation, initial-anchored target entropy, and dynamic global coefficient adjustment.
- 10
- 10
- CoSDA: Enhancing the Robustness of Inversion-based Generative Image Watermarking Framework9
A compensation-based forward sampling approach, compensating for certain mismatch conditions and reducing the inversion error caused by the mismatch, is proposed and Experimental results show that CoSDA effectively enhances watermark robustness while maintaining the visual quality of generated images.
- RoPaSS: Robust Watermarking for Partial Screen-Shooting Scenarios9
The RoPaSS is proposed: a robust watermarking scheme for partial screen-shooting scenarios, which effectively constructs symmetric characteristics on the embedding watermark to handle the sticky re-synchronization issue.
- DERO: Diffusion-Model-Erasure Robust Watermarking9
A destruction and compensation noise layer (DCNL) is designed to approximate the distortion effects caused by latent diffusion model erasure (LDE), which is significantly improved from 75% with SOTA methods to an impressive 96% with DERO.
- InferDPT: Privacy-Preserving Inference for Closed-box Large Language Model9
Experimental results indicate that PrivInfer is comparable to GPT-4 in text generation quality, and RANTEXT outperforms the current leading scheme in privacy protection, even under its adaptive attack, the proposed GPT inference attack.
- Prefix Guidance: A Steering Wheel for Large Language Models to Defend Against Jailbreak Attacks8
This paper proposes a plug-and-play and easy-to-deploy jailbreak defense framework, namely Prefix Guidance (PG), which guides the model to identify harmful prompts by directly setting the first few tokens of the model's output.
- Natias: Neuron Attribution-Based Transferable Image Adversarial Steganography8
A novel adversarial steganographic scheme named Natias, which attributes the output of a steganalytic model to each neuron in the target middle layer to identify critical features that may be adopted by diverse steganalytic models and can promote the transferability of adversarial steganography.
- Speech Pattern Based Black-Box Model Watermarking for Automatic Speech Recognition8
This paper proposes the first black-box model watermarking framework for protecting the IP of ASR models, which is robust against five kinds of attacks and has little impact on accuracy.
- Deflecting 3D Adversarial Point Clouds Through Outlier-Guided Removal.8
Simple random sampling (SRS) and statistical outlier removal (SOR) are proposed as defenses for 3D point cloud classification, where both methods remove points by estimating probability of points serving as adversarial points.
- Side Channel Steganalysis: When Behavior is Considered in Steganographer Detection8
This paper proposes utilizing the behavior of actors in the social network to identify the steganographer by utilizing the feature of subtractive images adjacent model (SIAM) to represent the correlations between images sequence of each actor.
- Multi-Label Text Classification with Label Attention Aware and Correlation Aware Contrastive Learning7
The Unified Contextual and Label-Aware Framework (UCLAF) is proposed, which integrates a Label Attention Aware Network (LAN) and Correlation Aware Contrastive Learning (CACL) in a synergistic design that enables UCLAF to jointly address complex label correlations and partial label overlaps.
- Performance-Lossless Black-Box Model Watermarking7
This work proposes a branch backdoor-based model watermarking protocol named BranchWM, which adopts a construction based on a message authentication scheme as the branch indicator, following a comparative analysis with other secure cryptographic primitives.
- Provably Secure Public-Key Steganography Based on Admissible Encoding7
The theoretical analysis and experimental results show that the proposed provable secure public-key steganography method can be deployed on all types of curves and utilize all points on the curve.
- GIFDL: Generated Image Fluctuation Distortion Learning for Enhancing Steganographic Security7
Inspired by the idea of natural steganography, GIFDL is proposed, a steganographic distortion learning method based on the fluctuations in generated images that takes a series of highly similar fluctuation images as the input to the steganographic distortion generator and introduces a new GAN training strategy to disguise stego images as fluctuation images.
- ProTegO: Protect Text Content against OCR Extraction Attack7
Experimental results demonstrate that ProTegO is a one-size-fits-all method that can ensure good visual quality while simultaneously achieving a high protection success rate on text recognition models with different architectures, outperforming the state-of-the-art methods.
- Controversial ‘pixel’ prior rule for JPEG adaptive steganography7
The authors develop a rule for cost reassignment in spatial domain called the `controversial pixel prior (CPP)' rule, which defines controversial pixels by utilizing the controversies among several comparable schemes and name it the J-CPP rule.
- Provably Secure Image Robust Steganography via Cross-modal Error Correction6
A high-quality, provably secure, and robust image steganography method based on state-of-the-art autoregressive (AR) image generation models using Vector-Quantized (VQ) tokenizers, which employs a cross-modal error-correction framework that generates stego text from stego images to aid in restoring lossy images, ultimately enabling the extraction of secret messages embedded within the images.
- 6
- 6
- Calibration-based Steganalysis for Neural Network Steganography6
A calibration-based steganalysis method is proposed, which fine-tunes the original neural network model without implicit constraints to obtain a reference model, then extracts and fuses statistical moments from the parameter distributions of the original model and itsreference model, and finally trains a logistic regressor for detection.
- 6
- 6
- Improving the Embedding Strategy for Batch Adaptive Steganography6
This paper proposes three methods for measuring the texture complexity of image to select images with highly textured content, then puts forward the universal embedding strategy for batch adaptive steganography in both spatial and JPEG domain, which is named max-residual-greedy (MRG) strategy.
- Targeted attack and security enhancement on texture synthesis based steganography6
Experiments demonstrate that the modified algorithm offers not only better resistance against the state-of-the-art steganalysis methods and steganalytic attack, but also a larger embedding capacity.
- Membership Inference Attacks on Tokenizers of Large Language Models5
The first study on membership leakage through tokenizers is presented and five attack methods to infer dataset membership are explored, highlighting tokenizers as an overlooked yet critical privacy threat and underscoring the urgent need for privacy-preserving mechanisms specifically designed for them.
- 5
- 5
- 5
- 5
- ICStega: Image Captioning-based Semantically Controllable Linguistic Steganography5
A novel image captioning-based stegosystem, where the secret messages are embedded into the generated captions and the semantics of the stego text can be controlled and the secret data can be transmitted by sending semantically related image-text pairs is put forward.
- Steganalysis of neural networks based on parameter statistical bias5
This work proposes steganalysis methods for steganography on neural network parameters that extract statistics from benign and malicious models and build classifiers based on the extracted statistics.
- 5
- 5
- SemBind: Binding Diffusion Watermarks to Semantics Against Black-Box Forgery Attacks4
Across four mainstream latent-based watermark methods, the SemBind-enabled anti-forgery variants markedly reduce false acceptance under black-box forgery while providing a controllable robustness-security balance.
- Breaking the Generative Steganography Trilemma: ANStega for Optimal Capacity, Efficiency, and Security4
ANStega introduces an innovative cryptographic mask mechanism that ensures the sampling process is driven by a cryptographically secure pseudo-random number generator, thereby achieving provable security and achieving exceptional embedding and extraction speeds.
- 4
- Provably Secure Robust Image Steganography via Cross-Modal Error Correction4
A high-quality, provably secure, and robust image steganography method based on state-of-the-art autoregressive (AR) image generation models using Vector-Quantized (VQ) tokenizers is proposed, which employs a cross-modal error-correction framework that generates stego text from stego images to aid in restoring lossy images, ultimately enabling the extraction of secret messages embedded within the images.
- 4
- LiteUpdate: A Lightweight Framework for Updating AI-Generated Image Detectors3
LiteUpdate, a lightweight framework for updating AI-generated image detectors to unseen generative models, employs a representative sample selection module that leverages image confidence and gradient-based discriminative features to precisely select boundary samples, significantly enhancing detector update efficiency.
- AI-generated image detection algorithm based on classical-quantum hybrid neural network3
This paper introduces a delicately designed QNN and combines it with the classical Swin Transformer V2, proposing an AI-generated image detection algorithm based on a classical-quantum hybrid neural network, and proposes an alternating layered ansatz (ALT)-based QNN, which successfully provides high trainability and rich expressibility.
- 3
- 3
- 3
- Provably Secure Steganography on Generative Media.3
It is proved theoretically that the proposed generative steganography framework using adaptive Arithmetic Coding is asymptotically perfect secure.
- 2
- 2
- WavInWav: Time-Domain Speech Hiding via Invertible Neural Network2
A flow-based invertible neural network is used to establish a direct link between stego audio, cover audio, and secret audio, enhancing the reversibility of embedding and extracting messages and exhibiting robustness to various types of noise, suggesting its utility in targeted secure communication scenarios.
- ReTokSync: Self-Synchronizing Tokenization Disambiguation for Generative Linguistic Steganography2
This work proposes ReTokSync (Re-Tokenization Synchronization), a self-synchronizing disambiguation framework that monitors the receiver-view tokenization during generation and triggers a corrective reset only when ambiguity actually occurs, and develops a two-channel covert communication mechanism.
- 2
- 2
- Clean Image May Be Dangerous: Data Poisoning Attacks Against Deep Hashing2
This work is the first to study data poisoning attacks against deep hashing (PADHASH), and it is pointed out that even clean query images can be dangerous, inducing malicious target retrieval results, like undesired or illegal images.
- Graph-Embedded Structure-Aware Perceptual Hashing for Neural Network Protection and Piracy Detection2
- 2
- 2
- 2
- 2
- 1
- SWIFT: Sliding Window Reconstruction for Few-Shot Training-Free Generated Video Attribution1
This work defines for the first time the "few-shot training-free generated video attribution" task and proposes SWIFT, which is tightly integrated with the temporal characteristics of the video, which achieves over 90% average attribution accuracy with merely 20 video samples across all models.
- DualCodeDetect: Zero-Shot LLM-Generated Code Detection via Dual-Channel Perturbation1
DualCodeDetect is proposed, a novel zero-shot detection framework that amplifies the differences between LLM-generated and human-written code through a dual-channel perturbation mechanism, and demonstrates strong cross-language generalization on Java, C++, and JavaScript.
- FARI: Robust One-Step Inversion for Watermarking in Diffusion Models1
FARI is proposed, a one-step inversion framework paired with lightweight adversarial LoRA fine-tuning of the denoiser for watermark extraction that surpasses 50-step DDIM inversion on watermark-verification robustness while dramatically reducing inference time.
- LD-RoViS: Training-free Robust Video Steganography for Deterministic Latent Diffusion Model1
LD-RoViS is proposed, a training-free and robust video steganography framework for the deterministic latent diffusion model that can embed approximately 12,000 bits of data into a 5-second video with an extraction accuracy exceeding 99%.
- MMPro: A Decoupled Perception-Thinking-Execution Framework for Secure GUI Agent1
By modularizing each stage while introducing a hybrid description language (HDL), MMPro transforms screen images into abstract structured representations, minimizing the risk of sensitive information leakage.
- 1
- 1
- StegoZip: Enhancing Linguistic Steganography Payload in Practice with Large Language Models1
StegoZip is a framework that leverages large language models to optimize secret message processing and achieves enhanced payload enhancement that significantly improves covertness by mitigating the risks associated with frequent transmissions while maintaining provable content security.
- AuthenLoRA: Entangling Stylization with Imperceptible Watermarks for Copyright-Secure LoRA Adapters1
AuthenLoRA is proposed, a unified watermarking framework that embeds imperceptible, traceable watermarks directly into the LoRA training process while preserving stylization quality and achieves high-fidelity stylization, robust watermark propagation, and significantly lower false-positive rates compared with existing approaches.
- 1
- Invertible Mask Network for Face Privacy-Preserving1
The proposed face privacy-preserving method based on Invertible"Mask"Network (IMN) can not only effectively protect the privacy of the protected face, but also almost perfectly recover theprotected face from the masked face.
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- WMVLM: Evaluating Diffusion Model Image Watermarking via Vision-Language Models–
WMVLM is proposed, the first unified and interpretable evaluation framework for diffusion model image watermarking via vision-language models (VLMs) and redefine quality and security metrics for each watermark type: residual watermarks are evaluated by artifact strength and erasure resistance, while semantic watermarks are assessed through latent distribution shifts.
- DNA: Dual-stage Native Attribution for Generated Image Source Tracing–
Dual-stage Native Attribution (DNA) is proposed, a coarse-to-fine framework that follows this hierarchy without additional neural-network training that achieves 89.11% end-to-end attribution accuracy on a task where random guessing accuracy is below 1% and outperforms the strongest baseline by 33.81% even when AEDR is used as the coarse-grained stage.
- –
- –
- Merging-Resistant Watermarking for LoRA Modules–
This paper proposes a novel merging-resistant watermarking method that embeds watermarks into the increment matrix generated during LoRA merging and decomposes the watermark-induced modifications into LoRA's standard matrices, achieving reliable watermark extraction and preserving LoRA's efficiency.
- Enhancing gig workers’ acceptance and mitigating algorithm aversion: impact of algorithmic transparency from the cognitive appraisal theory perspective–
The results show that perceived algorithmic transparency positively affects enjoyment and negatively affects fear, whereas fear mediates the relationship between perceived algorithmic transparency and aversion to the platform algorithm.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.