Article
Computer Science, Artificial Intelligence
Gaoming Yang, Mingwei Li, Xianjing Fang, Ji Zhang, Xingzhu Liang
Summary: Adversarial examples pose a security threat to deep learning models, and the proposed Attack Without a Target Model (AWTM) method achieves high attack success rate with low time cost.
PEERJ COMPUTER SCIENCE
(2021)
Article
Computer Science, Information Systems
Ghada Abdelmoumin, Jessica Whitaker, Danda B. Rawat, Abdul Rahman
Summary: The study highlights the limitations of current public datasets for training effective intelligent intrusion detection systems, emphasizing the importance of utilizing dynamically generated data in an adversarial setting. It suggests that training models using imbalanced and adversarial learning is crucial for enhancing the efficacy and performance of intrusion detection systems.
Article
Computer Science, Information Systems
Wen Ding Xiong, Kai Lun Luo, Rui Li
Summary: This paper proposes an Adversarial Intrusion Detection Training Framework (AIDTF) to improve the accuracy and robustness of IDS. It introduces an adversarial training method to obtain an IDS with high accuracy for both known test sets and unknown disguised attack samples.
COMPUTERS & SECURITY
(2023)
Article
Materials Science, Multidisciplinary
Teng Long, Yixuan Zhang, Nuno M. Fortunato, Chen Shen, Mian Dai, Hongbin Zhang
Summary: We developed an inverse design framework that enables automated generation of stable multicomponent crystal structures and discovered unreported crystal structures through analysis. This method provides convenience for inverse design of multicomponent materials with possible multi-objective optimization.
Article
Computer Science, Information Systems
Ye Peng, GuoBin Fu, Qi Yu, YingGuang Luo, Jia Hu, ChaoFan Duan
Summary: With the development of steganalysis technology, deep learning-based steganalyzers can accurately identify modification traces in steganographic covers, posing a significant threat to steganography. This research focuses on reducing the detection accuracy of deep learning-based steganalyzers. An Adversarial Example Steganography (AEST) method is designed to hide a secret grayscale image within a color cover image, creating a stego image that is difficult to distinguish. By using adversarial attacks such as FGM and PGD, small perturbations are added to generate adversarial steganographic images, effectively reducing the steganalyzer's detection accuracy. Additionally, a decoder based on adversarial training and generative adversarial networks is designed to minimize the impact of adversarial examples on secret information recovery. Experimental results demonstrate that AEST exhibits a strong anti-steganalysis performance, with the PGD attack-based adversarial steganographic image achieving a detection error rate of 63.511% for the XuNet steganalyzer.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Telecommunications
Junjun Chen, Di Wu, Ying Zhao, Nabin Sharma, Michael Blumenstein, Shui Yu
Summary: This paper introduces a novel attack framework AIDAE, which generates features to disable IDS and significantly degrades their detection performance. The framework uses an encoder and multiple decoders to transform features into a latent space, employs a generative adversarial network to learn a flexible prior distribution, and maintains the correlation between continuous and discrete features.
DIGITAL COMMUNICATIONS AND NETWORKS
(2021)
Article
Computer Science, Artificial Intelligence
Rongqian Zhang, Senlin Luo, Limin Pan, Jingwei Hao, Ji Zhang
Summary: In this paper, a novel method called NIDSFM is proposed for generating malicious traffic adversarial examples. The method reconstructs the feature space of traffic samples to isolate the discriminant features, avoiding interference with the malicious functions of the generated adversarial examples. The distribution of adversarial examples is modeled around the benign samples and fine-tuned using generative adversarial networks. Experimental results demonstrate that the proposed method significantly reduces the detection rate of multiple NIDSs and performs competitively in escaping NIDS detection.
Article
Computer Science, Information Systems
Mohammad Jamoos, Antonio M. Mora, Mohammad AlKhanafseh, Ola Surakhi
Summary: An intrusion detection system (IDS) is crucial for network security, and automated and intelligent IDSs are in high demand. Machine learning methods have proven effective in detecting malicious payloads in network traffic. However, the increasing volume of IDS data poses security risks and calls for stronger security measures. This paper proposes TDCGAN, a model-based generative adversarial network, to improve detection in imbalanced IDS datasets.
Article
Computer Science, Artificial Intelligence
Anibal Pedraza, Oscar Deniz, Gloria Bueno
Summary: The phenomenon of Adversarial Examples, where deep neural networks can be fooled by imperceptible perturbations, exists in the real world without maliciously selected noise. Through comparisons using distance and image quality metrics, it was shown that natural adversarial examples have a greater distance from the originals compared to artificially generated ones.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2022)
Article
Computer Science, Information Systems
Muhammad Shahzad Haroon, Husnain Mansoor Ali
Summary: Intrusion detection systems are crucial in network security, but machine learning-based systems are vulnerable to adversarial attacks. This study fills the gap in defense methods for network intrusion detection systems and conducts experiments using different machine learning algorithms.
CMC-COMPUTERS MATERIALS & CONTINUA
(2022)
Article
Computer Science, Information Systems
Paulo Freitas de Araujo-Filho, Georges Kaddoum, Divanilson R. Campelo, Aline Gondim Santos, David Macedo, Cleber Zanchettin
Summary: This article introduces a fog-based, unsupervised intrusion detection system (IDS) for CPSs using GANs, which calculates reconstruction loss based on data samples mapped to latent space for higher detection rates. Experimental results show that the proposed solution is at least 5.5 times faster than baseline methods.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Multidisciplinary Sciences
Ebtihaj Alshahrani, Daniyal Alghazzawi, Reem Alotaibi, Osama Rabie
Summary: The research conducted experiments on adversarial machine learning, indicating that evasion attacks had a significant impact on the accuracy of machine learning-based network intrusion detection systems.
Article
Chemistry, Analytical
Andrei-Grigore Mari, Daniel Zinca, Virgil Dobrota
Summary: Intrusion detection and prevention are crucial in network security infrastructure. Machine learning-based IDSs have been developed to detect malicious traffic that may evade traditional rules. This study focused on an IDS model using multiple algorithms and trained it using the NSL-KDD dataset. Adversarial instances of network traffic were created using a generative adversarial network (GAN) to test the IDS performance, and the results showed that using adversarial traffic improved the machine learning-based IDS performance even against traffic that could evade detection.
Article
Physics, Mathematical
Yang Zeng, Jin-Long Wu, Heng Xiao
Summary: This study enforces deterministic yet imprecise constraints on GANs by incorporating them into the loss function of the generator. The constrained GANs were found to produce samples that conform to the underlying constraints rather accurately on two representative tasks with geometrical and differential constraints.
COMMUNICATIONS IN COMPUTATIONAL PHYSICS
(2021)
Article
Computer Science, Information Systems
Phan The Duy, Nghi Hoang Khoa, Do Thi Thu Hien, Hien Do Hoang, Van-Hau Pham
Summary: Software Defined Networking (SDN) is used as the key technology to program and coordinate security policies in security operations centers (SOCs) for diverse networks. However, machine learning-based intrusion detection systems (ML-IDS) associated with SDN are susceptible to adversarial attacks due to the lack of diverse malicious records in the training dataset. This study explores the use of Wasserstein Generative Adversarial Networks (WGAN-GP), WGAN-GP with the two timescale update rule (WGAN-GP TTUR), and AdvGAN to generate perturbed attack samples for bypassing attack detectors and improving the resilience of ML-based IDSs in SDN.
JOURNAL OF INFORMATION SECURITY AND APPLICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Fenghui Zhang, Michael Mao Wang, Ruilong Deng, Xiaohu You
Summary: In this paper, a multi-agent independent learning approach is proposed to optimize the quality of service in the mobile ad hoc cloud. Simulation results confirm the effectiveness of the proposed approach.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2022)
Article
Automation & Control Systems
Chensheng Liu, Wangli He, Ruilong Deng, Yu-Chu Tian, Wenli Du
Summary: This article proposes a practical attack model for detecting malicious modification of critical network parameters. By exploiting the vulnerability of network parameter error processing and utilizing false-data-injection techniques, the requirements on attackers' capability and system information are significantly reduced. An optimal detection strategy is designed to minimize the number of protected measurements needed.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Theory & Methods
Zhenyong Zhang, Ruilong Deng, Youliang Tian, Peng Cheng, Jianfeng Ma
Summary: In this paper, a stealthy physics-manipulated attack (SPMA) is proposed by masking physical attacks on the flexible AC transmission system (FACTS) with strategic cyberattacks. The impact of physics manipulation on real-time economic dispatch and system operation security is analyzed. Countermeasures against SPMAs are also provided.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2023)
Article
Engineering, Electrical & Electronic
Zhenyong Zhang, Mingyang Sun, Ruilong Deng, Chongqing Kang, Mo-Yuen Chow
Summary: This paper proposes a novel concept called physics-constrained robustness, aiming to compute a lower-bound of adversarial perturbations for the ML-based intelligent security assessment (ISA) for power systems. Extensive experiments are conducted using real-world load profiles from New York State to evaluate the physics-constrained robustness of ISA in static and dynamic cases and provide suggestions for selecting ML models and parameters.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Mengxiang Liu, Chengcheng Zhao, Jinhui Xia, Ruilong Deng, Peng Cheng, Jiming Chen
Summary: In this paper, a proactive distributed detection and localization (PDDL) framework is proposed to defend against stealthy deception attacks in DC microgrids. Attack detection is achieved by observing voltage balancing deviation and current sharing deviation in DC microgrids. Once an anomaly is detected, a proactive perturbation is applied to the primary control gains to invalidate the inferred gains of the attacker, and the constructed stealthy deception attacks can be located using unknown input observer (UIO) based locators. An optimization problem is formulated to determine the magnitude of the perturbation to maximize attack locatability while limiting transient fluctuations on system states. The effectiveness of the PDDL framework is verified through hardware-in-the-loop (HIL) simulations and full-hardware experimental studies.
IEEE TRANSACTIONS ON SMART GRID
(2023)
Letter
Automation & Control Systems
Zhenyong Zhang, Yan Qin, Jingpei Wang, Hui Li, Ruilong Deng
Summary: In this letter, a detection method is proposed for the one-shot dummy attack (DA), a deep and stealthy data integrity attack that hides corrupted measurements in power industrial control processes. The method formulates an optimization problem to generate one-shot DAs, and then proposes an unsupervised data-driven approach based on a modified local outlier factor (MLOF) to detect them. Experimental results on real-world load data demonstrate the effectiveness of the proposed approach.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2023)
Letter
Automation & Control Systems
Zhenyong Zhang, Ruilong Deng
Summary: To enhance the security of the smart grid against false data injection attacks (FDIAs), the recently proposed moving target defense (MTD) strategy perturbs branch susceptances. However, previous research mainly focuses on the defending performance and the impact of MTD on static factors, neglecting the system dynamics. This letter studies the analytical impact of MTD on frequency stability, providing the condition for maintaining grid frequency stability and analyzing the effect of susceptance perturbation on frequency stability. The defending cost of MTD is also optimized considering defending performance and frequency stability constraints.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2023)
Article
Engineering, Electrical & Electronic
Lanting Zeng, Mingyang Sun, Xu Wan, Zhenyong Zhang, Ruilong Deng, Yan Xu
Summary: This paper proposes a physics-constrained vulnerability assessment framework for DRL-based power system operation and control, addressing the vulnerabilities and security threats. A novel adversarial example generation method is developed to conduct targeted adversarial attacks and evade bad data detection mechanisms. Case studies on the winners' models of the L2RPN competitions demonstrate the severe impacts on system operation and control.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2023)
Article
Automation & Control Systems
Zhenyong Zhang, Ruilong Deng, David K. Y. Yau, Peng Cheng, Mo-Yuen Chow
Summary: This article investigates the mechanism of Moving Target Defense (MTD) in power systems to counter False Data Injection Attacks (FDIAs) and develops an effective and low-cost MTD. A sufficient and necessary condition for protecting buses from intended FDIAs is provided, along with a new metric to quantify the level of protection and an efficient algorithm to minimize the number of required D-FACTS devices. Two strategies are proposed to reduce the operation cost of activating MTD, and the impact of MTD on system dynamics, particularly on small signal stability, is analyzed.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Editorial Material
Computer Science, Information Systems
Ruilong Deng, Chee-Wooi Ten, Chaojie Li, Dusit Niyato, Fei Teng
IEEE TRANSACTIONS ON CLOUD COMPUTING
(2023)
Article
Engineering, Electrical & Electronic
Shiyi Zhao, Jinhui Xia, Ruilong Deng, Peng Cheng, Qinmin Yang, Xuguo Jiao
Summary: The increasing use of wind energy necessitates attention to cybersecurity. This paper proposes a resilient torque control strategy for the wind turbine to counter denial-of-service attacks. By establishing mathematical models and designing a neural network observer, the proposed strategy can enhance attack resilience capability and ensure optimal rotor speed tracking.
IEEE TRANSACTIONS ON SMART GRID
(2023)
Article
Green & Sustainable Science & Technology
Shiyi Zhao, Jinhui Xia, Ruilong Deng, Peng Cheng, Qinmin Yang
Summary: With the increasing proportion of wind power generation in the power system, the vulnerabilities of the cyber links of wind turbines are gradually exposed. This paper develops an adaptive observer-based resilient control method to defend against time-delay attacks (TDA) in wind turbines. The proposed control scheme effectively mitigates the impact of TDA and ensures the output performance of the wind turbine system.
IEEE TRANSACTIONS ON SUSTAINABLE ENERGY
(2023)
Article
Computer Science, Information Systems
Zhenyong Zhang, Ke Zuo, Ruilong Deng, Fei Teng, Mingyang Sun
Summary: This article explores the vulnerability of machine learning-based intelligent systems enhanced with Internet of Things technologies in the stability assessment of electricity grids. It focuses on decision tree-based stability assessment approaches and investigates the feasibility of constructing a physics-constrained adversarial attack.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Computer Science, Theory & Methods
Hengye Zhu, Mengxiang Liu, Chongrong Fang, Ruilong Deng, Peng Cheng
Summary: This paper proposes an optimal watermarking design method for ICSs considering the tradeoff between detection performance. The watermark container is shifted from data points to segments, and detection metrics are updated to reduce noise interference. The proposed method is validated through numerical simulations and experiments.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2023)
Article
Automation & Control Systems
Chensheng Liu, Yang Tang, Ruilong Deng, Min Zhou, Wenli Du
Summary: Enabled by distributed flexible alternating current transmission system devices, MTD is considered an effective way to detect stealthy FDI attacks. However, limitations in power system topology prevent all attacks from being detected. In this article, the authors propose a joint MC-MTD method to improve detection by integrating MC with MTD. Theoretical analysis is performed on detection conditions and requirements, and an optimization is formulated to enhance effectiveness with low cost.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)