Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks34 from public data
- An active contour model for colour region extraction in natural scenes34
The proposed model is able to extract fairly complex object boundaries without the need to retune model parameters and image thresholds and is found to outperform the original snake model and degrade gracefully in the presence of image blur and Gaussian noise.
- A new colour image energy for active contours in natural scenes25
A new colour image energy for active contours is proposed to segment objects of unknown topologies automatically in the natural scene to handle the effects of uneven illumination and image blur in complex natural scenes.
- Transferable heterogeneous feature subspace learning for JPEG mismatched steganalysis23
A transferable heterogeneous feature subspace learning (THFSL) algorithm for JPEG mismatched steganalysis that can avoid a potentially negative transfer by using a sparse matrix to model the domain-related features and, thus, is more robust to different domain changes in mismatched Steganalysis.
- 22
- 21
- 21
- Reliable Model Watermarking: Defending against Theft without Compromising on Evasion20
The effectiveness of this method is demonstrated, showing not only improved robustness against evasion adversaries but also superior resistance to watermark removal attacks compared to state-of-the-art solutions.
- 19
- 18
- 12
- 12
- SelfPrompt: Autonomously Evaluating LLM Robustness via Domain-Constrained Knowledge Guidelines and Refined Adversarial Prompts11
A novel framework designed to autonomously evaluate the robustness of LLMs by incorporating refined adversarial prompts and domain-constrained knowledge guidelines in the form of knowledge graphs is introduced.
- An Adaptive JPEG Double Compression Steganographic Scheme Based on Irregular DCT Coefficients Distribution11
A secure JPEG double compression (DC) steganographic scheme based on irregular discrete cosine transformation (DCT) and coefficients distribution (IDCD-JDS), which can obtain less statistical detectability but performs better than the existing methods.
- 10
- Backdooring Self-Supervised Contrastive Learning by Noisy Alignment9
Noisy Alignment is proposed, a DPCL method that explicitly suppresses noise components in poisoned images, achieving state-of-the-art performance compared to existing DPCLs, while maintaining clean-data accuracy and demonstrating robustness against common backdoor defenses.
- Cross-Modal Unlearning via Influential Neuron Path Editing in Multimodal Large Language Models8
This work proposes a multimodal influential neuron path editor (MIP-Editor) for MU that introduces modality-specific attribution scores to identify influential neuron paths responsible for encoding forget-set knowledge and applies influential-path-aware neuron-editing via representation misdirection to enable effective and coordinated forgetting across modalities while preserving the model's general capabilities.
- 8
- Prompt as a Double-Edged Sword: A Dynamic Equilibrium Gradient-Assigned Attack against Graph Prompt Learning7
A dynamic equilibrium gradient-assigned attack against GPL, named MetaGpro, which outperforms the surrogate GNN in 96% of downstream evaluations, and reduces the accuracy of GPL by 2%∼20% compared to the state-of-the-art (SOTA) works mostly.
- Distributed normalisation input coding to speedup training processof BP-neural network classifier7
A coding method, distributed normalisation, is presented to speed up the training process of a back-propagation neural network classifier by distributing the values of the feature variables over a number of input nodes to increase the representation range of certain parts of each feature variable.
- Spoofing-aware Prompt Learning for Unified Physical-Digital Facial Attack Detection4
A Spoofing-aware Prompt Learning for Unified Attack Detection (SPL-UAD) framework, which decouples optimization branches for physical and digital attacks in the prompt space, enabling independent control of optimization for each attack type.
- PATFinger: Prompt-Adapted Transferable Fingerprinting against Unauthorized Multimodal Dataset Usage4
A novel prompt-adapted transferable fingerprinting scheme from a training-free perspective, called PATFinger, which incorporates the global optimal perturbation and the adaptive prompts to capture dataset-specific distribution characteristics and allows the dataset owner to check the usage of datasets by observing specific prediction behaviors linked to the PATFinger during retrieval queries.
- 4
- HCR-Auth: Reliable Bone Conduction Earphone Authentication with Head Contact Response4
This paper proposes HCR-Auth, a new authentication approach for bone conduction earphones based on head biometrics that employs a modulated chirp signal, emitted by the earphone speaker, to actively sense the user's head structure, and captures the response through the earphone's accelerometer.
- 3
- 3
- Distributed coding for data representation ofback-propagation neural network classifiers3
A new distributed input coding is derived by distributing the feature variables over a number of input nodes based on the distribution of the training data, which enables the network to converge at a higher rate during training.
- 2
- KGPA: Robustness Evaluation for Large Language Models via Cross-Domain Knowledge Graphs2
A framework that systematically evaluates the robustness of LLMs under adversarial attack scenarios by leveraging knowledge graphs (KGs) and generates original prompts from the triplets of knowledge graphs and creates adversarial prompts by poisoning, assessing the robustness of LLMs through the results of these adversarial attacks.
- 2
- 1
- –
- TGPA: Transferable graph prompt attack with hierarchical subgraph augmentation–
A novel transferable graph prompt attack, called TGPA, is proposed, which shifts the attack paradigm by introducing a hierarchical structural decoupling mechanism, which reduces the performance of pre-trained graph models with graph prompts by up to 28.9%, while guaranteeing robustness, stealthiness, and transferability.
- –
- A Secure and Robust Knowledge Transfer Framework via Stratified-Causality Distribution Adjustment in Intelligent Collaborative Services–
This work proposes a secure and robust knowledge transfer framework through stratified-causality distribution adjustment (SCDA) for device-edge-cloud collaborative services and introduces a cycle-consistent minimax optimization mechanism to ensure the effectiveness and dependability of knowledge transfer through the influence minimization and the diversity maximization.
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.