A5082961990- Anomaly Detection Techniques and Applications
- Advanced Malware Detection Techniques
- Network Security and Intrusion Detection
- Adversarial Robustness in Machine Learning
- Spam and Phishing Detection
- Bacillus and Francisella bacterial research
- Neuroscience and Neuropharmacology Research
- Pain Management and Opioid Use
- Radiomics and Machine Learning in Medical Imaging
- Integrated Circuits and Semiconductor Failure Analysis
- Anesthesia and Pain Management
- MRI in cancer diagnosis
- Face and Expression Recognition
- Risk and Safety Analysis
- Video Surveillance and Tracking Methods
- Physical Unclonable Functions (PUFs) and Hardware Security
- Biochemical and Molecular Research
- Mobile Ad Hoc Networks
- Glycosylation and Glycoproteins Research
- Electrical Fault Detection and Protection
- Advanced MIMO Systems Optimization
- Enhanced Recovery After Surgery
- Advanced Image and Video Retrieval Techniques
- Advanced Memory and Neural Computing
- Machine Learning and ELM
Institute of Biophysics
2022-2024
Chinese Academy of Sciences
2022-2024
University of Chinese Academy of Sciences
2023-2024
Fujian Provincial Hospital
2023
Chongqing University
2023
Fujian Medical University
2023
Yunnan Normal University
2011-2022
KDDI (Japan)
2019-2020
Kyushu University
2017-2019
Fujian Normal University
2018
Recent research has revealed that the output of Deep Neural Networks (DNN) can be easily altered by adding relatively small perturbations to input vector. In this paper, we analyze an attack in extremely limited scenario where only one pixel modified. For propose a novel method for generating one-pixel adversarial based on differential evolution (DE). It requires less information (a black-box attack) and fool more types networks due inherent features DE. The results show 67.97% natural...
The Internet of Things (IoT) is an extension the traditional Internet, which allows a very large number smart devices, such as home appliances, network cameras, sensors and controllers to connect one another share information improve user experiences. IoT devices are micro-computers for domain-specific computations rather than function-specific embedded devices. This opens possibility seeing many kinds existing attacks, traditionally targeted at also directed As shown by recent events, Mirai...
Abstract The output of convolutional neural networks (CNNs) has been shown to be discontinuous which can make the CNN image classifier vulnerable small well-tuned artificial perturbation. That is, images modified by conducting such alteration (i.e., adversarial perturbation) that little difference human eyes completely change classification results. In this paper, we propose a practical attack using differential evolution (DE) for generating effective perturbations. We comprehensively...
The aim of this study was to develop two nomograms for predicting pathologic complete response (pCR) after neoadjuvant chemotherapy (NACT) breast cancer based on quantitative dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), apparent diffusion coefficient (ADC), and clinicopathological characteristics at time-points: before cycles NACT, respectively.3.0 T MRI scans were performed 2 NACT in 215 patients. A total 74 female patients with stage II-III included. According univariate...
Convincing clinical evidence regarding completely opioid-free postoperative pain management using erector spinae plane block (ESPB) in patients undergoing open major hepatectomy (OMH) is lacking. Herein, we aimed to compare the analgesic efficacy of visualised continuous ESPB (VC-ESPB) and conventional intravenous opioid-based hepatocellular carcinoma (HCC) OMH.
Deep neural networks were shown to be vulnerable single pixel modifications. However, the reason behind such phenomena has never been elucidated. Here, we propose Propagation Maps which show influence of perturbation in each layer network. reveal that even extremely deep as Resnet, modification one easily propagates until last layer. In fact, this initial local is also spread becoming a global and reaching absolute difference values are close maximum value original feature maps given...
The Internet of Things (IoT) is an extension the traditional Internet, which allows a very large number smart devices, such as home appliances, network cameras, sensors and controllers to connect one another share information improve user experiences. Current IoT devices are typically micro-computers for domain-specific computations rather than functionspecific embedded devices. Therefore, many existing attacks, targeted at computers connected may also be directed For example, DDoS attacks...
Diffusion-weighted imaging (DWI) image quality will affect how well radiologists detect lesions and judge muscular invasion. This study qualitatively quantitatively compared the of DWI with integrated slice-specific dynamic shimming (iShim) single-shot echo-planar (SS-EPI) in diagnosis bladder cancer (BC) using 3.0 T magnetic resonance (MRI). We also investigated application value iShim BC.This retrospective enrolled 97 patients BC who underwent a preoperative MRI examination, including...
The malicious JavaScript is a common springboard for attackers to launch several types of network attacks, such as Drive-by-Download and PDF delivery attack. In order elude detection signature matching, often packed (so-called "obfuscation") with diversified algorithms therefore the occurrence obfuscation always good pointer potential maliciousness. this investigation, we propose light weight approach quickly filtering obfuscated by novel method tokenizing text at letter level...
Recent research has shown that deep convolutional neural networks (DCNN) are vulnerable to several different types of attacks while the reasons such vulnerability still under investigation. For instance, adversarial perturbations can conduct a slight change on natural image make target DCNN wrong recognition, is sensitive small modification divergent from one another. In this paper, we evaluate robustness two commonly used activation functions DCNN, namely sigmoid and ReLu, against recently...
The output of Convolutional Neural Networks (CNN) has been shown to be discontinuous which can make the CNN image classifier vulnerable small well-tuned artificial perturbations. That is, images modified by adding such perturbations(i.e. adversarial perturbations) that little difference human eyes, completely alter classification results. In this paper, we propose a practical attack using differential evolution(DE) for generating effective We comprehensively evaluate effectiveness different...
Abstract As an important part of power transmission and transformation equipment, vacuum circuit breakers have become the core device for protecting controlling electrical lines in high-voltage equipment. This paper expounds different types test methods preventive high low voltage distribution rooms. Taking 10kV system use as analysis object, 37 units are randomly selected samples, statistical principles used to analyze differences, extract point estimation interval Estimating eigenvalues,...
This paper proposes an adaptive rate strategy based on lora, which reduces the influence of crosstalk during data transmission and solves hidden terminal problem. Since spread factor, bandwidth coding lora technology can achieve different rates, channel state is evaluated by calculating received signal-to-noise ratio signal strength rts control frame fed back receiving end, determined according to threshold, so that adopt finally realize adaptation. Finally, algorithm simulated ns2, random...