A5100604099- Privacy-Preserving Technologies in Data
- Data Stream Mining Techniques
- Imbalanced Data Classification Techniques
- Machine Learning and Data Classification
- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Sentiment Analysis and Opinion Mining
- Advanced Statistical Methods and Models
- Text and Document Classification Technologies
- Food Supply Chain Traceability
- Target Tracking and Data Fusion in Sensor Networks
- Web Data Mining and Analysis
- Cryptography and Data Security
- Face recognition and analysis
- Advanced Algorithms and Applications
- Geophysical Methods and Applications
- Energy Load and Power Forecasting
- Smart Agriculture and AI
- Face and Expression Recognition
- Inertial Sensor and Navigation
- Privacy, Security, and Data Protection
- Time Series Analysis and Forecasting
- Advanced Malware Detection Techniques
- Water Systems and Optimization
- Solar Radiation and Photovoltaics
Xihua University
2024
China Southern Power Grid (China)
2023
Central China Normal University
2023
Hong Kong University of Science and Technology
2004-2022
University of Hong Kong
2004-2022
International University of the Caribbean
2021
University of Illinois Urbana-Champaign
2020
Concordia University
2019
East China Jiaotong University
2015
This paper studies model-inversion attacks, in which the access to a model is abused infer information about training data. Since its first introduction by~\cite{fredrikson2014privacy}, such attacks have raised serious concerns given that data usually contain privacy sensitive information. Thus far, successful only been demonstrated on simple models, as linear regression and logistic regression. Previous attempts invert neural networks, even ones with architectures, failed produce convincing...
In Machine Learning, the emergence of \textit{the right to be forgotten} gave birth a paradigm named \textit{machine unlearning}, which enables data holders proactively erase their from trained model. Existing machine unlearning techniques focus on centralized training, where access all holders' training is must for server conduct process. It remains largely underexplored about how achieve when full becomes unavailable. One noteworthy example Federated Learning (FL), each participating...
Clothing-change person re-identification (CC Re-ID) has attracted increasing attention in recent years due to its application prospect. Most existing works struggle adequately extract the ID-related information from original RGB images. In this paper, we propose an Identity-aware Feature Decoupling (IFD) learning framework mine identity-related features. Particularly, IFD exploits a dual stream architecture that consists of main and stream. The takes clothing-masked images as inputs derives...
Quantifying the importance of each training point to a learning task is fundamental problem in machine and estimated scores have been leveraged guide range data workflows such as summarization domain adaption. One simple idea use leave-one-out error indicate its importance. Recent work has also proposed Shapley value, it defines unique value distribution scheme that satisfies set appealing properties. However, calculating values often expensive, which limits applicability real-world...
As requirements for communication security grow, cryptographic processing becomes another type of application domain. However, algorithms are all computationally intensive. This work compares and analyzes architectural characteristics many widespread on the Intel IXP2800 network processor. It also investigates several implementation optimization principles that can improve overall performance. The results reported here applicable to other processors because they have similar components architectures.
As the requirements of smart, reliable and precise location for a vehicle, model fusion algorithm with big data selection accuracy correction is established to achieve low-cost location. In this paper, simple inertial navigation various positioning system sources different errors are intelligently selected, Kalman filtering used fuse information by function model, four kinds GPS, SINS, DR TDOA chosen simulate algorithm. The simulation results in MATLAB environment show that proposed...
The Internet of Things could benefit in several ways from mining data streams on connected objects rather than the cloud.In particular, limiting network communication with cloud services would improve user privacy and reduce energy consumption devices.Besides, applications leverage computing power for improved scalability.
Deep neural networks (DNNs) have been found to be vulnerable backdoor attacks, raising security concerns about their deployment in mission-critical applications. While existing defense methods demonstrated promising results, it is still not clear how effectively remove backdoor-associated neurons backdoored DNNs. In this paper, we propose a novel called \emph{Reconstructive Neuron Pruning} (RNP) expose and prune via an unlearning then recovering process. Specifically, RNP first unlearns the...
Things agriculture is an important development direction of networking applications, technology in the field for has brought immeasurable impetus and prospects immeasurable.In things, how can production process data collected processing, analysis display, making it better agricultural services needed to solve critical issues.Data processing located information application layer through integration scientific management decisions, achieve control.
Under cloudy weather, PV power will fluctuate dramatic due to the sporty cloud shading. However, most existing prediction models do not utilize map sufficiently. In this paper, we propose a novel method improve accuracy through combining features and hybrid neural network. Firstly, extract static dynamic of ground-based using image processing techniques. Secondly, establish an model based on ensemble empirical mode decomposition-bi-long short-term memory (EEMD-BiLSTM) with numerical weather...
Currently, the authenticity of historical data on photovoltaic power is compromised due to artificial restrictions and equipment failure during measurement communication. To address this issue ensure reliable follow-up research, paper proposes a method for identifying reconstructing outliers in with high outlier proportions. Initially, characteristics PV output its influencing factors are analyzed, leading classification types. Next, various basic identification models integrated using LSCP...