A5071801152- Network Security and Intrusion Detection
- Anomaly Detection Techniques and Applications
- Advanced Malware Detection Techniques
- IoT and Edge/Fog Computing
- Digital and Cyber Forensics
- Internet of Things and AI
- Advanced Computing and Algorithms
- Advanced Memory and Neural Computing
- Brain Tumor Detection and Classification
- Adversarial Robustness in Machine Learning
- Advanced Neural Network Applications
- Scientific Computing and Data Management
- Cloud Data Security Solutions
- Caching and Content Delivery
- Domain Adaptation and Few-Shot Learning
- Privacy-Preserving Technologies in Data
- Time Series Analysis and Forecasting
- Geophysical Methods and Applications
- Security and Verification in Computing
- Radiation Effects in Electronics
- Internet Traffic Analysis and Secure E-voting
China Academy of Engineering Physics
2022-2023
Harbin Institute of Technology
2021-2023
Chinese Academy of Engineering
2023
The appearance of container technology has profoundly changed the development and deployment multi-tier distributed applications. However, imperfect system resource isolation features kernel-sharing mechanism will introduce significant security risks to container-based cloud. In this paper, we propose a real-time unsupervised anomaly detection for monitoring calls in cloud via BiLSTM-based variational auto-encoder (VAE). Our proposed VAE network leverages generative characteristics learn...
Mobile edge computing (MEC) pushes resources to the of network and distributes them at mobile network. Offloading tasks instead cloud can reduce latency backhaul load simultaneously. However, new challenges incurred by user mobility limited coverage MEC server service arise. Services should be dynamically migrated between multiple servers maintain performance due movement. Tackling this problem is nontrivial because it arduous predict movement, migration will generate interruptions redundant...
Deep Learning as a Service (DLaaS) has become remarkable trend in modern data-driven online services.Both data holders and service providers need to build on trust thirdparty cloud infrastructure platforms.However, once the is broken, holders' sensitive providers' intellectual property rights will face significant security privacy risks.In this paper, we propose secure efficient inference framework for deep learning untrustworthy platforms, termed Branchy-TEE, which aims protect...
Existing black-box model inversion attacks mainly focus on training and attacking surrogate models. However, due to the deployment process of face recognition models, models becomes extremely difficult in practice. At same time, query-based still suffer from low image quality high computational costs. To bridge these gaps, this paper, we propose BMI-S, a sparse attack against BMI-S first introduces evolution strategies perform efficient gradient estimation achieve attacks. Meanwhile,...
The collaborative inference approach splits the Deep Neural Networks (DNNs) model into two parts. It runs collaboratively on end device and cloud server to minimize latency protect data privacy, especially in 5G era. scheme of DNN partitioning depends network bandwidth size. However, context dynamic mobile networks, resource-constrained devices cannot efficiently execute complex algorithms obtain optimal real-time. In this paper, overcome challenge, we first formulate problem as a Min-cut...
The Internet of Things (IoT) is widely used in industrial production and daily life, especially critical infrastructures. cybersecurity IoT has impacted national economic development, security, people's life safety. Deep learning been network security detection the IoT. However, traditional centralized approach can no longer meet demands regarding data privacy, load, timely response. We propose a collaborative cloud-edge-based intrusion to fill this gap. First, we reduce dimensionality based...
The cloud-edge collaborative inference approach splits deep neural networks (DNNs) into two parts to run collaboratively on resource-constrained edge devices and cloud servers, aiming at minimizing latency protecting data privacy for time-critical computing system. However, despite not exposing the raw input from directly cloud, state-of-the-art attacks can still target reconstruct private exposed local models' intermediate outputs, introducing serious risks.In this paper, we propose a...