A5100403360- Topic Modeling
- Security and Verification in Computing
- Advanced Text Analysis Techniques
- Cryptographic Implementations and Security
- Natural Language Processing Techniques
- Adversarial Robustness in Machine Learning
- Speech Recognition and Synthesis
- Anomaly Detection Techniques and Applications
- Advanced Malware Detection Techniques
- Physical Unclonable Functions (PUFs) and Hardware Security
- Advanced Data Storage Technologies
- Explainable Artificial Intelligence (XAI)
- Real-Time Systems Scheduling
- Technology and Security Systems
- Geochemistry and Geologic Mapping
- Covalent Organic Framework Applications
- Agricultural and Environmental Management
- Artificial Intelligence in Healthcare and Education
- Machine Learning in Healthcare
- Text and Document Classification Technologies
- Face and Expression Recognition
- Video Surveillance and Tracking Methods
- Parallel Computing and Optimization Techniques
- Sepsis Diagnosis and Treatment
- Geochemistry and Elemental Analysis
Max Planck Institute for Security and Privacy
2025
The University of Melbourne
2023-2024
State Grid Corporation of China (China)
2023
The University of Adelaide
2022
Fujian University of Technology
2021
Zhengzhou University of Light Industry
2020
Civil Aviation University of China
2015
China University of Geosciences (Beijing)
2014
Speculative out-of-order execution is a strategy of masking latency by allowing younger instructions to execute before older instructions. While originally considered be innocuous, speculative was brought into the spotlight with 2018 publication Spectre and Meltdown attacks. These attacks demonstrated that microarchitectural side channels can leak sensitive data accessed speculatively executed are not part normal program execution. Since then, significant effort has been vested in...
This paper introduces the design and implementation of a remote temperature data monitoring system based on STM32. The uses PT100 STM32F103VET6 main control boards to collect measured ambient temperature. Zigbee will pass tested through RS232 interface, deliver gateway MODBUS communication protocol. upload server MQTT protocol, receive for caching, storage analysis, display it web page. At same time, in order deal with problem network failure, InfluxDB standby database is deployed middle gateway.
The microarchitectural behavior of modern CPUs is mostly hidden from developers and users computer software. Due to a plethora attacks exploiting behavior, security-critical software must, e.g., ensure their code constant-time, which cumbersome usually results in slower programs. In practice, small leakages are deemed not exploitable still remain the codebase. For example, sub-cache-line have previously been investigated CacheBleed MemJam attacks, impractical on platforms.In this work, we...
Recent pretrained language models extend from millions to billions of parameters. Thus the need fine-tune an extremely large model with a limited training corpus arises in various downstream tasks. In this paper, we propose straightforward yet effective fine-tuning technique, Child-Tuning, which updates subset parameters (called child network) via strategically masking out gradients non-child network during backward process. Experiments on tasks GLUE benchmark show that Child-Tuning...
A topical text network is helpful when analyzing a large corpus of documents, for it can give an intuitive insight the topic distribution, including words and their connections. This paper proposes construction method based on seed word augmentation. Firstly, we manually select some representative each topic, then these are augmented by similarity metric. Secondly, defining threshold similarity, similar connected to construct network. We aviation safety reports analyze human factors using...
Pre-trained Language Models (PLMs) may be poisonous with backdoors or bias injected by the suspicious attacker during fine-tuning process. A core challenge of purifying potentially PLMs is precisely finding dimensions. To settle this issue, we propose Fine-purifying approach, which utilizes diffusion theory to study dynamic process for According relationship between parameter drifts and Hessians different dimensions, can detect dimensions abnormal dynamics, purify them resetting clean...
Deep Neural Networks (DNNs) are known to be vulnerable backdoor attacks. In Natural Language Processing (NLP), DNNs often backdoored during the fine-tuning process of a large-scale Pre-trained Model (PLM) with poisoned samples. Although clean weights PLMs readily available, existing methods have ignored this information in defending NLP models against work, we take first step exploit pre-trained (unfine-tuned) mitigate backdoors fine-tuned language models. Specifically, leverage via two...
Adversarial training is a method for enhancing neural networks to improve the robustness against adversarial examples. Besides security concerns of potential examples, can also generalization ability networks, train robust and provide interpretability networks. In this work, we introduce in time series analysis enhance better by taking finance field as an example. Rethinking existing research on training, propose adaptively scaled (ASAT) analysis, rescaling data at different slots with...
Abstract Background Establishing a mortality prediction model of patients undergoing cardiac surgery might be useful for clinicians alerting, judgment, and intervention, while few predictive tools long-term have been developed targeting post-cardiac surgery. Objective We aimed to construct validate several machine learning (ML) algorithms predict identify risk factors in unselected after during 4-year follow-up. Methods The Medical Information Mart Intensive Care (MIMIC-III) database was...