A5100697280- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Advanced Malware Detection Techniques
- Nuclear reactor physics and engineering
- Privacy-Preserving Technologies in Data
- Geophysical and Geoelectrical Methods
- Integrated Circuits and Semiconductor Failure Analysis
- Advanced Radiotherapy Techniques
- Non-Destructive Testing Techniques
- Innovative Energy Harvesting Technologies
- Medical Imaging Techniques and Applications
- Privacy, Security, and Data Protection
- Energy Harvesting in Wireless Networks
- Web Data Mining and Analysis
- Financial Distress and Bankruptcy Prediction
- Machine Learning in Healthcare
- Advanced Sensor and Energy Harvesting Materials
- Advanced Neural Network Applications
- Magnetic Field Sensors Techniques
- Mobile Crowdsensing and Crowdsourcing
- Geophysical Methods and Applications
- Target Tracking and Data Fusion in Sensor Networks
- Advanced Graph Neural Networks
- Radiation Dose and Imaging
- Wireless Power Transfer Systems
Harbin Institute of Technology
2019-2025
Chongqing Medical University
2025
Taiyuan University of Technology
2024
Visa (United Kingdom)
2021-2024
Northwestern Polytechnical University
2023-2024
Fudan University
2023
Tsinghua University
2018-2023
Hainan Normal University
2022-2023
Central University of Finance and Economics
2023
Northwest University of Politics and Law
2023
Many modern databases include personal and sensitive correlated data, such as private information on users connected together in a social network, measurements of physical activity single subjects across time. However, differential privacy, the current gold standard data does not adequately address privacy issues this kind data.
Motivated by safety-critical applications, test-time attacks on classifiers via adversarial examples has recently received a great deal of attention. However, there is general lack understanding why arise; whether they originate due to inherent properties data or training samples remains ill-understood. In this work, we introduce theoretical framework analogous bias-variance theory for these effects. We use our analyze the robustness canonical non-parametric classifier - k-nearest neighbors....
We consider data poisoning attacks, a class of adversarial attacks on machine learning where an adversary has the power to alter small fraction training in order make trained classifier satisfy certain objectives. While there been much prior work poisoning, most it is offline setting, and for online learning, arrives streaming manner, are not well understood. In this work, we initiate systematic investigation learning. formalize problem into two settings, propose general attack strategy,...
Detection of adversarial examples with high accuracy is critical for the security deployed deep neural network-based models. We present first graph-based detection method that constructs a Latent Neighborhood Graph (LNG) around an input example to determine if adversarial. Given example, selected reference and benign (represented as LNG nodes in Figure 1) are used capture local manifold vicinity example. The node connectivity parameters optimized jointly graph attention network end-to-end...
Film dosimetry is commonly performed by using linear CCD array transmission optical densitometers. However, these devices suffer from a variation in response along the detector array. If not properly corrected for, this nonuniformity may lead to significant overestimations of measured dose as one approaches regions close edges scanning region. In note, we present measurements spatial an AGFA Arcus II document scanner used for radiochromic film dosimetry. Results and methods presented work...
Airborne magnetic anomaly detection is an important passive remote sensing technique. However, since the field caused by aircraft interferes with accuracy, this part of interference should be eliminated aeromagnetic compensation method. Most existing methods assume that ambient uniform when calculating model parameters. as actually not and varies location, solved parameters ignore related to varied field. Although some latest deep learning-based avoid assumption uniformity field,...
Federated Learning (FL) enables collaborative model training while keeping client data private. However, exposing individual updates makes FL vulnerable to reconstruction attacks. Secure aggregation mitigates such privacy risks but prevents the server from verifying validity of each update, creating a privacy-robustness tradeoff. Recent efforts attempt address this tradeoff by enforcing checks on using zero-knowledge proofs, they support limited predicates and often depend public validation...
A serious game module called "D-Casting" was developed in the previous study. This study aimed to determine effectiveness of D-Casting module. The experiment consisted two parts: construct validity assessment and skill transfer assessment. Eligible participants, who were students majoring dental technology, recruited from Stomatology College Chongqing Medical University. intervention designed based on games framework. a total 145 participants (100% response rate). results suggested that...
Abstract Magnetic anomaly detection (MAD) is a technique to find ferromagnets hiding in strong and complicated magnetic background. In many practical cases, the targets are very far from sensor, which leads low signal‐to‐noise ratio (SNR) high difficulty. Most of current methods determine existence target by some approaches based on signal analysis, such as orthogonal basis function (OBF) minimum entropy (ME). However, although these consume resources, performances not satisfactory enough....
Solar-driven interfacial evaporation is an efficient method for purifying contaminated or saline water. Nonetheless, the suboptimal design of structure and composition still necessitates a compromise between rate service life. Therefore, achieving production clean water remains key challenge. Here, biomimetic dictyophora hydrogel based on loofah/carbonized sucrose@ZIF-8/polyvinyl alcohol demonstrated, which can serve as independent solar evaporator recovery. This special structural achieves...
Magnetic anomaly detection (MAD) is a method that uses magnetometers to find hidden ferromagnetic objects based on variations in magnetic field signals. Many methods use time–frequency fusion features detect target signals, but these are extracted using unlearnable the preprocessing. This article proposes network with adaptive feature expression anomalies. The adaptively learns of signal through optimization, and step, it selects key for classification. In experiments, we compared our...
Abstract Dangerous driving, e.g., using mobile phone while can result in serious traffic problem and threaten to safety. To efficiently alleviate such problem, this paper, we design an intelligent monitoring system detect the dangerous behavior driving. The is combined by a designed target detection algorithm, camera, terminal server voice reminder. An deep learning model, namely Mobilenet with single shot multi-box detector (Mobilenet-SSD), was applied identify of driver. evaluate...
In recent years, there has been a growing interest in the detection, location, and classification (DLC) of multiple dipole-like magnetic sources based on gradient tensor (MGT) data. these applications, tilt angle is usually used to detect number sources. We found that only suitable for scenario where positive negative signs sources' inclination are same. Therefore, we map L2 norm vertical arctan function, denoted as VMGT2 angle, Then use normalized source strength (NSS) narrow parameters'...
Toward robust malware detection, we explore the attack surface of existing detection systems. We conduct root-cause analyses practical binary-level black-box adversarial examples. Additionally, uncover sensitivity volatile features within engines and exhibit their exploitability. Highlighting information channels software, introduce three software pre-processing steps to eliminate surface, namely, padding removal, stripping, inter-section resetting. Further, counter emerging section...
Network public opinion refers to the common with tendency and influence formed by on certain social events through Internet. Due complexity of interest relations, network is likely cause difficulties for individuals, enterprises or governments. In order control public's emotional events, this study designed an OCC sentiment rule system label case base. The text representation method Word2Vec in deep learning, convolution neural used construct analysis model under opinion. Taking Dujia Banna...