A5100324043- Privacy-Preserving Technologies in Data
- Privacy, Security, and Data Protection
- Cryptography and Data Security
- Imbalanced Data Classification Techniques
- Blockchain Technology Applications and Security
- Cloud Data Security Solutions
- Internet Traffic Analysis and Secure E-voting
- Natural Language Processing Techniques
- Image Retrieval and Classification Techniques
- Health disparities and outcomes
- IoT and Edge/Fog Computing
- Human Pose and Action Recognition
- Computational Drug Discovery Methods
- Anomaly Detection Techniques and Applications
- Ethics in Clinical Research
- Multimodal Machine Learning Applications
- Advanced Computational Techniques and Applications
- Crime, Illicit Activities, and Governance
- Advanced Neural Network Applications
- Adversarial Robustness in Machine Learning
- Gait Recognition and Analysis
- Vehicle License Plate Recognition
- Artificial Intelligence in Healthcare and Education
- Advanced Image and Video Retrieval Techniques
- Biomedical Text Mining and Ontologies
Swinburne University of Technology
2021-2025
Xiamen University of Technology
2024
The University of Texas Health Science Center at Houston
2024
Lehigh University
2022
Wuhan University of Science and Technology
2019
NARI Group (China)
2017
Blockchain has gradually attracted widespread attention from the research community of IoT, due to its decentralization, consistency, and other attributes. It builds a secure robust system by generating backup locally for each participant node collectively maintain network. However, this feature brings some privacy concerns since all nodes can access chain data, users’ sensitive information under risk leakage. The local differential (LDP) mechanism be promising way address issue as it...
Notary cross-chain transaction technologies have obtained broad affirmation from industry and academia as they can avoid data islands enhance chain interoperability. However, the increased privacy concern in sharing makes participants hesitate to upload sensitive information without trust foundation of external network. To address this issue, paper proposes a differential private notary mechanism (DPNM) preserve blockchain interoperations. It establishes fully trusted organization conduct...
The post-processed Differential Privacy (DP) framework has been routinely adopted to preserve privacy while maintaining important invariant characteristics of datasets in data-release applications such as census data. Typical include non-negative counts and total population. Subspace DP proposed population guaranteeing for sub-populations. Non-negativity post-processing identified inherently incur fairness issues. In this work, we study unfairness ( <italic...
Federated learning (FL) can be essential in knowledge representation, reasoning, and data mining applications over multi-source graphs (KGs). A recent study FedE first proposes an FL framework that shares entity embeddings of KGs across all clients. However, embedding sharing from would incur a severe privacy leakage. Specifically, the known used to infer whether specific relation between two entities exists private client. In this paper, we introduce novel attack method aims recover...
Applying sparsity- and overfitting-aware eXtreme Gradient Boosting (XGBoost) for classification in federated learning allows many participants to train a series of trees collaboratively. Since various local multiclass distributions global aggregation diversity, model performance plummets as convergence slowly accuracy decreases. Worse still, neither the nor server can detect this problem make timely adjustments. In paper, we provide new local-global class imbalance inconsistency...
The cross-modal molecule retrieval (Text2Mol) task aims to bridge the semantic gap between molecules and natural language descriptions. A solution this nontrivial problem relies on a graph convolutional network (GCN) attention with contrastive learning for reasonable results. However, there exist following issues. First, mechanism is only in favor of text representations cannot provide helpful information representations. Second, GCN-based encoder ignores edge features importance various...
In recent years, money laundering has become much easier to be achieved but more challenging detected than before, which enormous adversary effects on finance, military, and other related fields. the real-time scenario, every case a unique structure in terms of transactions. It is not sufficient detect suspicious behavior by just following probability theory, where usually thresholds are given experts. Since crime prevalent sophisticated nowadays, it will increase complexity detection if...
A number of solutions have been proposed to tackle the user privacy-preserving issue. Most existing schemes, however, focus on methodology and techniques from perspective data processing. In this paper, we propose a lightweight scheme for identity applied cryptography. The basic idea is break association relationships between User his behaviors ensure that can access or services as usual while real will not be revealed. To end, an interactive zero-knowledge proof protocol executed CSP User....
Federated learning is a promising paradigm that allows multiple clients to collaboratively train model without sharing the local data. However, presence of heterogeneous devices in federated learning, such as mobile phones and IoT with varying memory capabilities, would limit scale hence performance could be trained. The mainstream approaches address limitations focus on width-slimming techniques, where different subnetworks reduced widths locally then server aggregates subnetworks. global...
Credit card defaults cost the economy tens of billions dollars every year. However, financial institutions rarely collaborate to build more comprehensive models due legal regulations and competition. Federated XGBoost is an emerging paradigm that enables several companies a classification model cooperatively without transferring local data others. The conventional suffers from inverse inference according splitting nodes selection class imbalance problem severely. Utilising characteristic...
This paper introduces FedSecurity, an end-to-end benchmark designed to simulate adversarial attacks and corresponding defense mechanisms in Federated Learning (FL). FedSecurity comprises two pivotal components: FedAttacker, which facilitates the simulation of a variety during FL training, FedDefender, implements defensive counteract these attacks. As open-source library, enhances its usability compared from-scratch implementations that focus on specific attack/defense scenarios based...
In the platform of supervisory control system based on SOA, message bus provides facilities data exchange between different processes in computers as an important middleware. With continuous expansion application scope and scale, quantity via increases rapidly. It is urgent to improve performance provide better quality service. Furthermore, new generation carries out partitions according function or region into cooperative subsystems for flexibility. Message needs be redesigned meet...
Federated learning (FL) can be essential in knowledge representation, reasoning, and data mining applications over multi-source graphs (KGs). A recent study FedE first proposes an FL framework that shares entity embeddings of KGs across all clients. However, embedding sharing from would incur a severe privacy leakage. Specifically, the known used to infer whether specific relation between two entities exists private client. In this paper, we introduce novel attack method aims recover...
We consider a federated representation learning framework, where with the assistance of central server, group $N$ distributed clients train collaboratively over their private data, for representations (or embeddings) set entities (e.g., users in social network). Under this key step aggregating local embeddings trained privately at clients, we develop secure embedding aggregation protocol named \scheme, which leverages all potential opportunities among while providing privacy guarantees and...
As a very popular framework, federated learning can help heterogeneous participants cooperate training global models without the local data being exposed. It not only takes advantage of massive raw data, but also fundamentally protects privacy participants. An unavoidable challenge is that class imbalance brought by many will seriously affect model performance and even damage convergence. Introducing Focal loss to dynamically adjust weight samples in process good choice for relieving this...
Abstract To resolve the communication overhead problem of anonymous users, we propose a location privacy protection method based on cache technology. In particular, first place center edge server nodes to reduce interaction between servers and users. this way, risk leaks can be reduced. Furthermore, improve caching hit rate, prediction system Markov chain is designed protect trajectory mobile Simulations show that algorithm users transmission delay.