A5100333272- Network Security and Intrusion Detection
- Advanced Malware Detection Techniques
- Internet Traffic Analysis and Secure E-voting
- Adversarial Robustness in Machine Learning
- Advanced Algorithms and Applications
- Anomaly Detection Techniques and Applications
- Spam and Phishing Detection
- Privacy-Preserving Technologies in Data
- Software-Defined Networks and 5G
- Advanced Computational Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Complex Network Analysis Techniques
- Network Packet Processing and Optimization
- Industrial Technology and Control Systems
- Neural Networks and Applications
- Digital Media Forensic Detection
- Combustion and flame dynamics
- Smart Grid and Power Systems
- Topic Modeling
- Cryptography and Data Security
- Advanced Sensor and Control Systems
- Advanced Graph Neural Networks
- Text and Document Classification Technologies
- Advanced Combustion Engine Technologies
- Security and Verification in Computing
First Automotive Works (China)
2024
INTI International University
2023-2024
Northwest University
2010-2024
China University of Geosciences
2019-2024
Xinzhou Teachers University
2012-2023
Indiana University Bloomington
2009-2023
Shanghai Urban Construction Design and Research Institute (Group)
2023
PLA Air Force Aviation University
2019-2022
National University of Defense Technology
2009-2022
Huaqiao University
2021
With more IoT devices entering the consumer market, it becomes imperative to detect their security vulnerabilities before an attacker does.Existing binary analysis based approaches only work on firmware, which is less accessible except for those equipped with special tools extracting code from device.To address this challenge in analysis, we present paper a novel automatic fuzzing framework, called IOTFUZZER, aims at finding memory corruption without access firmware images.The key idea upon...
Recently, many studies have demonstrated deep neural network (DNN) classifiers can be fooled by the adversarial example, which is crafted via introducing some perturbations into an original sample. Accordingly, powerful defense techniques were proposed. However, existing often require modifying target model or depend on prior knowledge of attacks. In this paper, we propose a straightforward method for detecting image examples, directly deployed unmodified off-the-shelf DNN models. We...
Membership Inference Attack (MIA) determines the presence of a record in machine learning model's training data by querying model. Prior work has shown that attack is feasible when model overfitted to its or adversary controls algorithm. However, not and does control algorithm, threat well understood. In this paper, we report study discovers overfitting be sufficient but necessary condition for an MIA succeed. More specifically, demonstrate even well-generalized contains vulnerable instances...
Membership Inference Attacks (MIAs) aim to determine the presence of a record in machine learning model's training data by querying model. Recent work has demonstrated effectiveness MIA on various models and corresponding defenses have been proposed. However, both attacks focused an adversary that indiscriminately all records without regard cost false positives or negatives. In this work, we revisit membership inference from perspective pragmatic who carefully selects targets make...
We present <i>congestion puzzles</i> (CP), a new countermeasure to bandwidth-exhaustion attacks. Like other defenses based on client puzzles, CP attempts force attackers invest vast resources in order effectively perform denial-of-service Unlike previous puzzle-based approaches, however, ours is the first designed for attacks that are common at network (IP) layer. At core of an elegant distributed puzzle mechanism permits routers cooperatively impose and check puzzles. demonstrate through...
Network servers and applications commonly use static IP addresses communication ports, making themselves easy targets for network reconnaissances attacks. Port address hopping is a novel effective moving target defense (MTD) which hides by constantly changing ports. In this paper, we develop port mechanism called Random Address Hopping (RPAH), unpredictably mutates ports based on source identity, service identity as well time with high rate. RPAH provides us more strength MTD three...
The issue of multi-step-ahead time series prediction is a daunting challenge predictive modeling. In this work, we propose multi-output iterative model with stacking LSTM neural network (MO-LSTMs). the proposed model, utilize that consists multiple hidden layers to learn features data, and use dropout algorithm improve generalization ability robustness deep learning method. our network, each layer contains different units memory state cells in are reset. method solves problem single...
BulletProof Hosting (BPH) services provide criminal actors with technical infrastructure that is resilient to complaints of illicit activities, which serves as a basic building block for streamlining numerous types attacks. Anecdotal reports have highlighted an emerging trend these BPH reselling from lower end service providers (hosting ISPs, cloud hosting, and CDNs) instead monolithic providers. This has rendered many the prior methods detecting less effective, since being highly...
Take-down operations aim to disrupt cybercrime involving malicious domains.In the past decade, many successful take-down have been reported, including those against Conficker worm, and most recently, VPNFilter.Although it plays an important role in fighting cybercrime, domain procedure is still surprisingly opaque.There seems be no in-depth understanding about how operation works whether there due diligence ensure its security reliability.In this paper, we report first systematic study on...
Neural text ranking models have witnessed significant advancement and are increasingly being deployed in practice. Unfortunately, they also inherit adversarial vulnerabilities of general neural models, which been detected but remain underexplored by prior studies. Moreover, the might be leveraged blackhat SEO to defeat better-protected search engines. In this study, we propose an imitation attack on black-box passage models. We first show that target model can transparentized imitated...
A security threat to deep neural networks (DNN) is backdoor contamination, in which an adversary poisons the training data of a target model inject Trojan so that images carrying specific trigger will always be classified into label. Prior research on this problem assumes dominance image's representation, causes any image with recognized as member class. Such also exhibits unique features representation space and can therefore easily separated from legitimate images. Our research, however,...
Face detection is a preliminary step for wide range of applications such as face recognition, video-surveillance and so on. The object this work to improve the correct rate detection. Skin-color model established first extract possible region, then BP(Back Propagation) neural network used simulate output human region Bayesian decision theory classify or non-face pattern. Experiments show that color images using skin-color effective fast; use BP simulation remove dummy distinguish an way;...