Article
Computer Science, Artificial Intelligence
Shuyi Yang, Mattia Cerrato, Dino Ienco, Ruggero G. Pensa, Roberto Esposito
Summary: Semi-supervised learning shows potential in real-world applications with few labeled examples available, but struggles with fairness constraints. To address this, we propose a fair semi-supervised representation learning architecture that achieves fair and accurate results even with limited biased instances. Experimental results show the competitiveness of our approach in general and real-world settings.
Article
Engineering, Electrical & Electronic
Maryam Farajzadeh-Zanjani, Ehsan Hallaji, Roozbeh Razavi-Far, Mehrdad Saif, Masood Parvania
Summary: This paper introduces a novel adversarial scheme for semi-supervised learning to address the issues of partially labelled samples and skewed class distributions. By leveraging the generator and discriminator in an adversarial manner, more synthetic minority class samples are generated to train the model to learn the distribution of minority class samples, resulting in superior performance in diagnosing attacks and faults.
IEEE TRANSACTIONS ON SMART GRID
(2021)
Article
Computer Science, Information Systems
Hongzuo Xu, Yijie Wang, Guansong Pang, Songlei Jian, Ning Liu, Yongjun Wang
Summary: Semi-supervised anomaly detection methods improve performance compared to unsupervised models by using a few anomaly examples. However, they have limitations regarding anomaly contamination and suboptimal learning of anomaly scores. This paper proposes a novel method that addresses these limitations by devising contamination-resilient continuous supervisory signals, which significantly outperforms existing methods in various settings.
INFORMATION PROCESSING & MANAGEMENT
(2023)
Article
Computer Science, Information Systems
Ines Ortega-Fernandez, Marta Sestelo, Juan C. Burguillo, Camilo Pinon-Blanco
Summary: In this paper, we propose a network intrusion detection system (NIDS) architecture based on a deep autoencoder trained on network flow data. The experimental results show that the proposed model can detect anomalies caused by distributed denial of service attacks, providing a high detection rate and low false alarms, outperforming the state-of-the-art and a baseline model.
Article
Automation & Control Systems
Guangdou Zhang, Jian Li, Olusola Bamisile, Yankai Xing, Di Cao, Qi Huang
Summary: This study proposes an automatic identification and classification method for multiple cyber-attacks based on the deep capsule convolution neural network. The proposed method extracts spatial correlations and temporal features from the history operation status in the transmitted data packets. Numerical results show that the proposed method achieves high detection accuracy on both single cyber-attacks and multiple cyber-attacks.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Construction & Building Technology
Mostafa Mohammadpourfard, Abdullah Khalili, Istemihan Genc, Charalambos Konstantinou
Summary: Future cities face a challenge in achieving environmental sustainability while also defending against cyber threats. Ensuring the cybersecurity of smart grids is crucial, especially with growing uncertainties and the impact of renewable energy resources. The development of LSTM-based attack detection models shows promise in accurately capturing the dynamic behaviors of modern power grids and outperforming traditional methods in detecting real-time attacks.
SUSTAINABLE CITIES AND SOCIETY
(2021)
Article
Biochemical Research Methods
Zile Wang, Haiyun Wang, Jianping Zhao, Chunhou Zheng
Summary: This study proposes a semi-supervised clustering model called scSemiAAE for scRNA sequence analysis. Through the use of deep generative neural networks, scSemiAAE significantly improves clustering performance compared to other unsupervised and semi-supervised algorithms, promoting downstream analysis.
BMC BIOINFORMATICS
(2023)
Article
Telecommunications
Mahmoud Said El Sayed, Nhien-An Le-Khac, Marianne A. Azer, Anca D. Jurcut
Summary: Software Defined Networking (SDN) is an emerging network platform that enables centralised network management. However, it also brings new security concerns, such as Distributed Denial of Service (DDoS) attacks. This paper proposes using feature selection methods and deep learning techniques to tackle DDoS attacks in SDN networks.
IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING
(2022)
Article
Physics, Multidisciplinary
Jie Lai, Xiaodan Wang, Qian Xiang, Wen Quan, Yafei Song
Summary: The efficiency and cognitive limitations of manual sample labeling lead to a large number of unlabeled training samples. Making use of both labeled and unlabeled samples is crucial for solving the semi-supervised problem. The study introduces the pseudo-labeling method into the stacked autoencoder (SAE) and proposes a novel pseudo label-based semi-supervised stacked autoencoder (PL-SSAE) to improve the performance of semi-supervised classification tasks.
Article
Energy & Fuels
Urtzi Otamendi, Inigo Martinez, Marco Quartulli, Igor G. Olaizola, Elisabeth Viles, Werther Cambarau
Summary: This article introduces an end-to-end deep learning pipeline that utilizes deep learning techniques to detect, locate, and segment cell-level anomalies in photovoltaic modules via EL images. The modular pipeline combines object detection, image classification, and weakly supervised segmentation techniques, allowing for upgrades and extensions towards further improvements and new functionalities in the state-of-the-art.
Article
Computer Science, Cybernetics
Kamal Berahmand, Yuefeng Li, Yue Xu
Summary: Network clustering is an unsupervised method that aims to group similar nodes together. Semi-supervised clustering detection, which utilizes side information, is a promising approach for community detection. To address the limitations of previous methods, we propose an end-to-end deep semi-supervisor community detection (DSSC) method.
IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xiangli Yang, Zixing Song, Irwin King, Zenglin Xu
Summary: This paper provides a comprehensive survey on deep semi-supervised learning methods, including model design and unsupervised loss functions. It categorizes existing methods into different types and reviews 60 representative methods with a detailed comparison. The paper also discusses the shortcomings of existing methods and proposes heuristic solutions.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Engineering, Civil
Yuanzhe Wang, Qipeng Liu, Ehsan Mihankhah, Chen Lv, Danwei Wang
Summary: This paper explores the cybersecurity challenges faced by autonomous vehicles when under sensor attacks. A model-based framework is proposed to detect attacks and identify their sources for secure localization. Sensor redundancy and a bank of attack detectors are utilized to ensure robustness against cyber-attacks.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Computer Science, Information Systems
Stephanie Harshbarger, Mohsen Hosseinzadehtaher, Alireza Zare, Amin Y. Fard, Mohammad B. Shadmand, George Amariucai
Summary: This paper investigates the impact of stealthy zero-dynamics attacks on power electronic dominated grids (PEDGs) and highlights their higher destructive potential compared to traditional power systems. By exploiting the low inertia characteristic of PEDGs, attackers can cause frequency instability, resulting in more harm to the system in a shorter time. The increasing number of telecommunication devices in PEDGs also provides attackers with a larger attack surface. The study emphasizes the importance of designing PEDGs with measures to minimize susceptibility to stealthy attacks.
Article
Computer Science, Information Systems
Xiaoxi Zhang, Yuan Gao, Xin Wang, Jun Feng, Yan Shi
Summary: This paper investigates the problem of transportation mode identification using GPS trajectories and geographic information, and proposes a geographic information-fused semi-supervised method. The proposed method can train an excellent transportation mode identification model with only a few labeled samples.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2022)
Article
Chemistry, Multidisciplinary
Qingqing Li, Penghui Dong, Jun Zheng
APPLIED SCIENCES-BASEL
(2020)
Article
Computer Science, Information Systems
Anhao Xiang, Jun Zheng
Article
Chemistry, Multidisciplinary
Zhirui Luo, Qingqing Li, Jun Zheng
Summary: This study investigates adversarial attacks and detection on DL-based plant disease identification systems, revealing that attacks with a small number of perturbations can significantly degrade DNN model performance. It also proposes defense mechanisms against adversarial attacks, providing a basis for developing more robust DNN models and guiding defense strategies in plant disease identification.
APPLIED SCIENCES-BASEL
(2021)
Article
Chemistry, Analytical
Mathew Salas, Sihua Shao, Adrian Salustri, Zachary Schroeck, Jun Zheng
Summary: Smart appliances and electric vehicles can be managed through a smart grid enabled home area network (HAN), which helps to reduce electricity demand during critical times and decrease greenhouse gas emissions. However, the commonly used wireless communication technologies in HAN are vulnerable to cyber attacks. Therefore, a low-cost solution using retro-reflector based visible light communication (Retro-VLC) is proposed for secure data exchange.
Article
Computer Science, Information Systems
Michelle Sherman, Sihua Shao, Xiang Sun, Jun Zheng
Summary: In urban environments, tall buildings or structures can limit direct wireless communication between a base station (BS) and an Internet of Things device (IoTD). The use of unmanned aerial vehicles (UAVs) with reconfigurable intelligent surfaces (RIS) as relay nodes, known as UAV-RIS, has been proposed to enhance the system throughput capacity in wireless access networks. Uncoordinated UAVs or RIS phase shift elements can negatively impact signal transmission to IoTDs. To minimize Average Sum of Age of Information (ASoA), two model-free deep reinforcement learning (DRL) approaches, Off-Policy deep Q-network (DQN) and On-Policy proximal policy optimization (PPO), were developed to optimize RIS phase shift, UAV-RIS location, and IoTD transmission scheduling in large-scale Internet of Things wireless networks. Analysis and simulations show the superiority of the On-Policy approach, PPO, in terms of stability, convergence speed, and adaptability to different environments.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Proceedings Paper
Green & Sustainable Science & Technology
Elijah Orozco, Ruobin Qi, Jun Zheng
Summary: Cyber attacks on smart grids' Advanced Metering Infrastructure (AMI), such as electricity theft, can lead to significant financial losses for utility companies. In this study, we propose a semi-supervised outlier detection approach for data-driven electricity theft detection, utilizing only normal energy usage data for training. We investigate nine semi-supervised outlier detection algorithms and perform feature engineering to extract 20 time-series features from energy load profiles, aiding in the detection of malicious changes in load curve or energy usage. The results demonstrate that the proposed feature engineering significantly improves the performance of semi-supervised outlier detection algorithms.
2023 IEEE GREEN TECHNOLOGIES CONFERENCE, GREENTECH
(2023)
Article
Engineering, Electrical & Electronic
Ruobin Qi, Jun Zheng, Zhirui Luo, Qingqing Li
Summary: This article proposes a novel method for detecting electricity theft in AMI, which uses observer meter data, wavelet-based feature extraction, and fuzzy c-means clustering to differentiate normal and fraudulent users. Experimental results show that the proposed method achieves significantly better performance compared to existing methods.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Engineering, Electrical & Electronic
Qingqing Li, Zhirui Luo, Jun Zheng
Summary: This article investigates the use of multichannel sEMG signals of hand gestures for user authentication. A new deep anomaly detection-based method is proposed, which employs sEMG images generated from multichannel sEMG signals. Among different sEMG image generation methods, the root mean square (rms) map achieves the best performance.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Computer Science, Information Systems
Zhirui Luo, Qingqing Li, Jun Zheng
Summary: The increasing popularity of social media has made it easier to create and spread rumors, which could cause devastating damages. This paper introduces a new deep feature fusion method for Twitter rumor detection, utilizing linguistic characteristics and propagation tree patterns to achieve significantly better detection performance than other baseline methods.
Proceedings Paper
Computer Science, Information Systems
Andrew Corum, Donovan Jenkins, Jun Zheng
2019 2ND INTERNATIONAL CONFERENCE ON DATA INTELLIGENCE AND SECURITY (ICDIS 2019)
(2019)
Article
Computer Science, Information Systems
Jun Zheng, Sai Krishna Chigurupati
Proceedings Paper
Engineering, Electrical & Electronic
Geetika Kovelamudi, Jun Zheng, Srinivas Mukkamala
2016 IEEE/CIC INTERNATIONAL CONFERENCE ON COMMUNICATIONS IN CHINA (ICCC)
(2016)