Ziyao Wang

Ziyao Wang

Auto-generated profile, based on publicly available information

Ziyao Wang

University of Maryland College Park

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 December 2024
Verified
Conference Paper
Open access
PWPeiran Wang
Peiran Wang
QLQiyu Li
Qiyu Li
LYLongxuan Yu
Longxuan Yu
ZWZiyao Wang
Ziyao Wang
ALANG Li
ANG Li
HJHaojian Jin
Haojian Jin
We present Moderator, a policy-based model management system that allows administrators to specify fine-grained content moderation policies and modify the weights of a text-to-image (TTI) model to make it significantly more challenging for users to produce images that violate the policies. In contrast to existing general-purpose model editing techniques, which unlearn concepts without considering the associated contexts, Moderator allows admins to specify what content should be moderated, under which context, how it should be moderated, and why moderation is necessary. Given a set of policies, Moderator first prompts the original model to generate images that need to be moderated, then uses these self-generated images to reverse fine-tune the model to compute task vectors for moderation and finally negates the original model with the task vectors to decrease its performance in generating moderated content. We evaluated Moderator with 14 participants to play the role of admins and found they could quickly learn and author policies to pass unit tests in approximately 2.29 policy iterations. Our experiment with 32 stable diffusion users suggested that Moderator can prevent 65% of users from generating moderated content under 15 attempts and require the remaining users an average of 8.3 times more attempts to generate undesired content.

30 April 2023
Verified
Conference Paper
Open access
XGXueluan Gong
Xueluan Gong
ZWZiyao Wang
Ziyao Wang
YCYanjiao Chen
Yanjiao Chen
QWQian Wang
Qian Wang
CWCong Wang
Cong Wang
SCShen Chao
Shen Chao
Recently more and more cloud service providers (e.g., Microsoft, Google, and Amazon) have commercialized their well-trained deep learning models by providing limited access via web API interfaces. However, it is shown that these APIs are susceptible to model inversion attacks, where attackers can recover the training data with high fidelity, which may cause serious privacy leakage.Existing defenses against model inversion attacks, however, hinder the model performance and are ineffective for more advanced attacks, e.g., Mirror [4]. In this paper, we proposed NetGuard, a novel utility-aware defense methodology against model inversion attacks (MIAs). Unlike previous works that perturb prediction outputs of the victim model, we propose to mislead the MIA effort by inserting engineered fake samples during the training process. A generative adversarial network (GAN) is carefully built to construct fake training samples to mislead the attack model without degrading the performance of the victim model. Besides, we adopt continual learning to further improve the utility of the victim model. Extensive experiments on CelebA, VGG-Face, and VGG-Face2 datasets show that NetGuard is superior to existing defenses, including DP [37] and Ad-mi [32] on state-of-the-art model inversion attacks, i.e., DMI [8], Mirror [4], Privacy [12], and Alignment [34].

2 December 2024
Verified
Conference Paper
Open access
XGXueluan Gong
Xueluan Gong
RWRubin Wei
Rubin Wei
ZWZiyao Wang
Ziyao Wang
YSYuchen Sun
Yuchen Sun
JPJiawen Peng
Jiawen Peng
QWQian Wang
Qian Wang
Machine Learning as a Service (MLaaS) enables resource-constrained users to access well-trained models through a publicly accessible Application Programming Interface (API) on a pay-per-query basis. Nevertheless, model owners may face the potential threats of model extraction attacks where malicious users replicate valuable commercial models based on query results. Existing defenses against model extraction attacks, however, either sacrifice prediction accuracy or fail to thwart more advanced attacks. In this paper, we propose a novel model extraction defense, dubbed Beowulf 1, which draws inspiration from theoretical findings that models with complex and narrow decision regions are difficult to be reproduced. Rather than arbitrarily altering decision regions, which may jeopardize the predictive capacity of the victim model, we introduce a dummy class, carefully synthesized using both random and adversarial noises. The random noise broadens the coverage of the dummy class, and the adversarial noise impacts decision regions near decision boundaries with normal classes. To further improve the model utility, we propose to employ data augmentation methods to seamlessly integrate the dummy class and the normal classes. Extensive evaluations on CIFAR-10, GTSRB, CIFAR-100, and ImageNette datasets demonstrate that Beowulf can significantly reduce the extraction accuracy of 6 state-of-the-art model extraction attacks by as much as 80%. Moreover, we show that Beowulf is also robust to adaptive model extraction attacks.

1 January 2023
Verified
Conference Paper
Open access
PDF file
ZWZiyao Wang
Ziyao Wang
HVHoang Thien Van
Hoang Thien Van
DLDongwon Lee
Dongwon Lee
Ziyao Wang, Thai Le, Dongwon Lee. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.

1 March 2025 in arXiv
Verified
Preprint
Open access
PDF file
YHYexiao He
Yexiao He
ZWZiyao Wang
Ziyao Wang
YZYuning Zhang
Yuning Zhang
Tingting DanTingting Dan
Tingting Dan
Tianlong ChenTianlong Chen
Tianlong Chen
ALAng Li
Ang Li
Alzheimer's disease (AD) diagnosis is complex, requiring the integration of imaging and clinical data for accurate assessment. While deep learning has shown promise in brain MRI analysis, it often functions as a black box, limiting interpretability and lacking mechanisms to effectively integrate critical clinical data such as biomarkers, medical history, and demographic information. To bridge this gap, we propose NeuroSymAD, a neuro-symbolic framework that synergizes neural networks with symbolic reasoning. A neural network percepts brain MRI scans, while a large language model (LLM) distills medical rules to guide a symbolic system in reasoning over biomarkers and medical history. This structured integration enhances both diagnostic accuracy and explainability. Experiments on the ADNI dataset demonstrate that NeuroSymAD outperforms state-of-the-art methods by up to 2.91% in accuracy and 3.43% in F1-score while providing transparent and interpretable diagnosis.

31 January 2026 in arXiv
Verified
Preprint
Open access
PDF file
ZWZiyao Wang
Ziyao Wang
DJDaeun Jung
Daeun Jung
YHYexiao He
Yexiao He
GSGuoheng Sun
Guoheng Sun
ZSZheyu Shen
Zheyu Shen
ALANG Li
ANG Li
Group Relative Policy Optimization (GRPO) has recently emerged as an effective approach for improving the reasoning capabilities of large language models through online multi-objective reinforcement learning. While personalization on private data is increasingly vital, traditional Reinforcement Learning (RL) alignment is often memory-prohibitive for on-device federated learning due to the overhead of maintaining a separate critic network. GRPO's critic-free architecture enables feasible on-device training, yet transitioning to a federated setting introduces systemic challenges: heterogeneous reward definitions, imbalanced multi-objective optimization, and high training costs. We propose FedMOA, a federated GRPO framework for multi-objective alignment under heterogeneous rewards. FedMOA stabilizes local training through an online adaptive weighting mechanism via hypergradient descent, which prioritizes primary reasoning as auxiliary objectives saturate. On the server side, it utilizes a task- and accuracy-aware aggregation strategy to prioritize high-quality updates. Experiments on mathematical reasoning and code generation benchmarks demonstrate that FedMOA consistently outperforms federated averaging, achieving accuracy gains of up to 2.2% while improving global performance, personalization, and multi-objective balance.