A5100393772- Adversarial Robustness in Machine Learning
- Topic Modeling
- Natural Language Processing Techniques
- Anomaly Detection Techniques and Applications
- Security and Verification in Computing
- Text and Document Classification Technologies
- Advanced Malware Detection Techniques
- Speech Recognition and Synthesis
- Climate variability and models
- Advanced Neural Network Applications
- Sentiment Analysis and Opinion Mining
- Solar Radiation and Photovoltaics
- IoT and Edge/Fog Computing
- Advanced Text Analysis Techniques
- Meteorological Phenomena and Simulations
- Advanced Steganography and Watermarking Techniques
- Sarcoma Diagnosis and Treatment
- Atmospheric chemistry and aerosols
- Rangeland Management and Livestock Ecology
- Cryptographic Implementations and Security
- Radiation Effects in Electronics
- Rangeland and Wildlife Management
- Handwritten Text Recognition Techniques
- Distributed Control Multi-Agent Systems
- Digital Marketing and Social Media
Minzu University of China
2024
Third Hospital of Hebei Medical University
2024
Hebei Medical University
2024
Shanghai Maritime University
2020-2024
Guangdong University of Technology
2022-2024
Nanjing Xiaozhuang University
2023
Changchun Institute of Technology
2023
Tencent (China)
2022
Beijing Normal University
2019-2021
Hefei University of Technology
2020
Deep Neural Networks (DNNs) have recently led to significant improvements in many fields. However, DNNs are vulnerable adversarial examples which samples with imperceptible perturbations while dramatically misleading the DNNs. Moreover, can be used perform an attack on various kinds of DNN based systems, even if adversary has no access underlying model. Many defense methods been proposed, such as obfuscating gradients networks or detecting examples. However it is proved out that these not...
Deep neural network based steganalysis has developed rapidly in recent years, which poses a challenge to the security of steganography. However, there is no steganography method that can effectively resist networks for at present. In this paper, we propose new strategy constructs enhanced covers against with technique adversarial examples. The and their corresponding stegos are most likely be judged as by networks. Besides, use both deep high-dimensional feature classifiers evaluate...
Deep neural networks have recently achieved tremendous success in image classification. Recent studies however shown that they are easily misled into incorrect classification decisions by adversarial examples. Adversaries can even craft attacks querying the model black-box settings, where no information about is released except its final decision. Such decision-based usually require lots of queries, while real-world recognition systems might actually restrict number queries. In this paper,...
Online image sharing in social platforms can lead to undesired privacy disclosure. For example, some enterprises may detect these large volumes of uploaded images do users’ in-depth preference analysis for commercial purposes. And their technology might be today’s most powerful learning model, deep neural network (DNN). To just elude automatic DNN detectors without affecting visual quality human eyes, we design and implement a novel Stealth algorithm , which makes the detector blind...
Abstract Trusted execution environments (TEE) are deployed on many platforms to provide both confidentiality and integrity, their extensive use offers a secure environment for privacy-sensitive operations. Despite TEE prevalence in the smartphone tablet market, vulnerability research into security is relatively rare. This is, part, due strong isolation guarantees provided by its implementation. In this paper, we propose hardware assisted fuzzing framework, CROWBAR, that bypasses natively...
Deep neural networks (DNNs) have achieved tremendous success in many tasks of machine learning, such as the image classification. Unfortunately, researchers shown that DNNs are easily attacked by adversarial examples, slightly perturbed images which can mislead to give incorrect classification results. Such attack has seriously hampered deployment DNN systems areas where security or safety requirements strict, autonomous cars, face recognition, malware detection. Defensive distillation is a...
This paper describes our work in participation of the IWSLT-2021 offline speech translation task. Our system was built a cascade form, including speaker diarization module, an Automatic Speech Recognition (ASR) module and Machine Translation (MT) module. We directly use LIUM SpkDiarization tool as The ASR is trained with three datasets from different sources, by multi-source training, using modified Transformer encoder. MT pretrained on large-scale WMT news dataset fine-tuned TED corpus....
On E-commerce stores, there are rich recommendation content to help shoppers shopping more efficiently. However given numerous products, it's crucial select most relevant reduce the burden of information overload. We introduced a ranking service powered by linear causal bandit algorithm rank and for each shopper under context. The mainly leverages aggregated customer behavior features, ignores single level past activities. study problem inferring interest from historical propose deep...
Due to the large number of attributes and low repetition rate in a cross-lingual knowledge graph, it is difficult for an alignment task embed attribute information efficiently. To solve problem, entity model based on weight updating network was proposed. Firstly, order efficiently, embedding approximately constructed with through constructor, thus avoiding their separate training. Secondly, fact that different make contributions alignment, module graph attention proposed update each using...
Abstract In this article, a modified mosaic approach method (MMAM), which considers the subgrid‐scale effect of topographical height on atmospheric forcing, is introduced. Two experiments are designed to study effects MMAM surface meteorological variables in Western China within weather research and forecasting (WRF) modelling framework during June August 2010. Results show that has obvious 2 m temperature improvement. Simulations with closer observed values over wide range areas,...
To satisfy the low-latency requirement of vehicular applications, partial offloading strategy for multiple tasks in edge computing (VEC) was studied. This fully utilized resources roadside units (RSUs) and nearby vehicles to reduce total task processing latency. A VEC multitask scheme based on an improved particle swarm optimization (PSO) algorithm proposed, which consisted three steps: first, under latency constraint resource constraint, optimal ratio each local, determined; second, within...
In Internet of Things (IoT) endpoint devices, some data or address ciphers are used for real-time memory protection to mitigate side-channel attacks against memories. To better meet the requirements protection, this article proposes a hardware engine permutation-based encryption (PAE) implement with flexible width adaptation, low latency, and overhead. When evaluated TSMC's 40-nm standard CMOS technology, PAE features lightweight characteristics gate count 0.589 KGates, which is only 0.37%...
Trusted Execution Environments (TEEs) are deployed in many CPU designs because of the confidentiality and integrity guarantees they provide. ARM TrustZone is a TEE extensively on smart phones, IoT devices, notebooks. Specifically, used to separate code execution data into two worlds, normal world secure world. However, this separation inherently prevents traditional fuzzing approaches which rely upon coverage-guided feedback existing research is, therefore, extremely limited. In paper, we...
Chinese Spelling Check (CSC) aims to detect and correct spelling errors in Chinese. Most CSC models rely on human-defined confusion sets narrow the search space, failing resolve outside set. However, most current benchmark datasets are character pairs similar pronunciations. Errors shapes which visually phonologically irrelevant not considered. Furthermore, widely-used automatically generated training data tasks leads label leakage unfair comparison between different methods. In this work,...