A5100668775- Adversarial Robustness in Machine Learning
- Privacy-Preserving Technologies in Data
- Cryptography and Data Security
- Integrated Circuits and Semiconductor Failure Analysis
- Graphene and Nanomaterials Applications
- Domain Adaptation and Few-Shot Learning
- Blockchain Technology Applications and Security
- Anomaly Detection Techniques and Applications
- Imbalanced Data Classification Techniques
- Smart Grid Energy Management
- Face recognition and analysis
- Generative Adversarial Networks and Image Synthesis
- Library Science and Information Systems
- Stochastic Gradient Optimization Techniques
- Frequency Control in Power Systems
- Advanced Algorithms and Applications
- Advanced Sensor and Control Systems
- Facial Trauma and Fracture Management
- Industrial Technology and Control Systems
- Microgrid Control and Optimization
- Sports and Physical Education Research
- Transportation and Mobility Innovations
- Image Processing and 3D Reconstruction
- Electric Vehicles and Infrastructure
- Topic Modeling
NARI Group (China)
2024
Zhoukou Normal University
2024
Tencent (China)
2022
Cloud Computing Center
2020
Federated learning (FL), as a distributed machine setting, has received considerable attention in recent years. To alleviate privacy concerns, FL essentially promises that multiple parties jointly train the model by exchanging gradients rather than raw data. However, intrinsic issue still exists FL, e.g., user's training samples could be revealed solely inferring gradients. Moreover, emerging poisoning attack also poses crucial security threat to FL. In particular, due nature of malicious...
Fog-cloud computing promises many new vertical service areas beyond simple data communication, storing, and processing. Among them, distributed deep learning (DDL) across fog-cloud environment is one of the most popular applications due to its high efficiency scalability. Compared with centralized learning, DDL can provide better privacy protection training only on sharing parameters. Nevertheless, when meets computing, it still faces two major security challenges: 1) how protect users' from...
The effectiveness of state-of-the-art deep learning (DL) models has empowered the development industrial Internet things (IIoT). Recently, considering resource-constrained and privacy-required IIoT devices, collaborative inference been proposed, which splits DL deploys them in devices an edge server separately. However, this article, we argue that there are still severe privacy vulnerabilities systems. And devise first membership attack (MIA) against inference, to infer whether a particular...
Federated learning (FL) has emerged as a powerful technology widely applied in Internet of Things (IoT). Recently, researchers have shown an increased interest privacy-preserving FL with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">unreliable users</i> . The goal such works is to achieve private training under ciphertext mode while ensuring that the model mainly derived from contributions users high-quality data. However, existing work...
The popularity of electric vehicles depends on convenient and efficient charging services. At present, none existing services allow users to reach stations at desirable time charge immediately when they arrive without waiting. This paper proposes a reservation service approach based the consortium blockchain smart contract technology. Users can choose station period with no congestion, which is information recorded in composed located distributed regions city. To ensure user arrives within...
Based on the theory of system engineering, coal mine production safety has been analyzed. From workers, process, and economy, evaluation indexes have confirmed, then BP neural network model built. After training testing model, can accurately evaluate safety. It is great guiding significance for
With the increasing number of financial transactions, fraud has become increasingly serious for institutions and public. The core idea this model is to integrate multiple neural network structures utilize their respective advantages improve performance detection. Firstly, we employed convolutional with interpretable blocks (CNNIB) (CNN) extract key features from data capture patterns in cases. Secondly, introduced autoencoder generative adversarial (AE-GAN) perform feature analysis on...
Oracle bone inscriptions(OBI) is the earliest developed writing system in China, bearing invaluable written exemplifications of early Shang history and paleography. However, task deciphering OBI, current climate scholarship, can prove extremely challenging. Out 4,500 oracle characters excavated, only a third have been successfully identified. Therefore, leveraging advantages advanced AI technology to assist decipherment OBI highly essential research topic. fully utilizing AI's capabilities...
As the complexity of microgrid systems, randomness load disturbances, and data dimensionality increase, traditional frequency control methods for microgrids are no longer capable handling such highly complex nonlinear systems. This can result in this significant fluctuations oscillations, potentially leading to blackouts microgrids. To address random power disturbances introduced by a large amount renewable energy, paper proposes Learning-Driven Load Frequency Control (LD-LFC) method....
A hybrid model of Convolutional Neural Network (CNN) and self-attention has achieved remarkable results in text classification fields. In previous researches, local semantics(captured by CNN) global representation (extracted self-attention) play equally important roles for each input. However, the importance two varies greatly with complex linguistic backgrounds. this paper, we take an adaptive approach to automatically determine contribution degree classification, according specific...
We propose HifiHead, a high fidelity neural talking head synthesis method, which can well preserve the source image's appearance and control motion (e.g., pose, expression, gaze) flexibly with 3D morphable face models (3DMMs) parameters derived from driving image or indicated by users. Existing works mainly focus on low-resolution inputs. Instead, we exploit powerful generative prior embedded in StyleGAN to achieve high-quality editing. Specifically, first extract construct descriptors, are...