Robust Detection Research Articles

Robust fault detection and isolation for uncertain neutral time-delay systems using a geometric approach

This paper proposes a new geometric fault detection and isolation (FDI) strategy for uncertain neutral time-delay systems (UNTDS). Firstly, the concept of unobservability subspace is extended to the considered system. Subsequently, utilizing the geometric properties of factor space and canonical projection, the fault is divided into different unobservability subspaces. Therefore, an algorithm for constructing the subspace is developed for fault isolation. Finally, a set of observers is designed for the subsystems, and generates a set of structured residuals which is sensitive only to a specific fault. Additionally, the H∞ technique is utilized to suppress the disturbances and error signals due to time-varying delays on the residual. The simulation examples verify the effectiveness of the proposed approach.

Security Analysis for Smart Healthcare Systems.

The healthcare industry went through reformation by integrating the Internet of Medical Things (IoMT) to enable data harnessing by transmission mediums from different devices, about patients to healthcare staff devices, for further analysis through cloud-based servers for proper diagnosis of patients, yielding efficient and accurate results. However, IoMT technology is accompanied by a set of drawbacks in terms of security risks and vulnerabilities, such as violating and exposing patients' sensitive and confidential data. Further, the network traffic data is prone to interception attacks caused by a wireless type of communication and alteration of data, which could cause unwanted outcomes. The advocated scheme provides insight into a robust Intrusion Detection System (IDS) for IoMT networks. It leverages a honeypot to divert attackers away from critical systems, reducing the attack surface. Additionally, the IDS employs an ensemble method combining Logistic Regression and K-Nearest Neighbor algorithms. This approach harnesses the strengths of both algorithms to improve attack detection accuracy and robustness. This work analyzes the impact, performance, accuracy, and precision outcomes of the used model on two IoMT-related datasets which contain multiple attack types such as Man-In-The-Middle (MITM), Data Injection, and Distributed Denial of Services (DDOS). The yielded results showed that the proposed ensemble method was effective in detecting intrusion attempts and classifying them as attacks or normal network traffic, with a high accuracy of 92.5% for the first dataset and 99.54% for the second dataset and a precision of 96.74% for the first dataset and 99.228% for the second dataset.

Analyzing group polarization through text emotion measurement and time series prediction: A comparative study across three online platforms

This study investigated the emotional trends of users on social platforms, considering the event of “the Changsha girl jumping off the Lalamove truck” as a case study. It examined the effects of recommendation algorithms and group social comparison attributes on group emotions across three platforms: Zhihu, Weibo, and Bilibili. Through text mining and emotion analysis algorithms, group reviews were analyzed, and an event-based ARIMA robustness detection model was constructed using time series data. Utilizing the theoretical framework of the social comparison process, the study discovered that the “information database” formed by the recommendation algorithms of social platforms fosters the emergence of emotional group polarization among users. Furthermore, the findings revealed that the audience's social comparison attributes play a role in shaping emotional group polarization. High knowledge attributes tend to inhibit emotional group polarization, while low knowledge attributes tend to promote it. Machine learning algorithms were employed to measure user sentiment in social media platform comments, revealing insights into the causes of group polarization through a comparison of social comparison attributes and algorithm techniques. Future studies must focus on measuring technical information entropy.

Domain-Adaptive and Context-Aware Fall Detection Based on Coarse-Fine Network Learning

Accurate fall detection among older adults is crucial for minimizing injuries and fatalities. However, existing fall detection systems face challenges due to the rarity and variability of falls, compounded by limitations in real-world datasets. To address this, a novel fall detection approach integrating domain adaptation and context-awareness within a Coarse-Fine Network Learning framework is proposed. The model combines high-level semantic understanding with low-level spatial details to achieve robust fall detection across diverse environments. Domain adaptation techniques like transfer learning and domain-specific fine-tuning are introduced to enhance model generalization and adaptability. Additionally, context-aware features, including environmental cues and behavioral patterns, reduce false alarms. Extensive experimentation on real- world datasets demonstrates the superior performance of the model, outperforming traditional approaches. The framework holds promise for deployment in healthcare settings, contributing to improved safety for older adults worldwide. The interpretability of the model's predictions enhances its usability in practical applications.

CAIMP: Cross-Architecture IoT Malware Detection and Prediction Based On Static Feature

Abstract IoT malware and cross-platform malware are currently the top threats to information systems. This paper proposes a robust cross-architecture IoT malware detection and prediction model based on machine learning and opcode features using a novel approach. In our method, a feature opcode transformation model between chip architecture platforms is proposed to facilitate the process of building a detection model for cross-architecture malware on IoT devices. The feature transformation model is capable of converting opcodes between different architecture platforms using an unsupervised machine learning approach. In our approach, a machine learning model is used for the detection of cross-platform malware based on the proposed opcode features. Experiments have demonstrated that our method is effective in detecting and predicting cross-platform malware with an accuracy of up to 99.4% and an F1-score of 99.3%. The method is capable of learning on one architecture platform and detecting malware on a different architecture platform. Therefore, the method can be used to develop cross-architecture detection and zero-day malware prediction solutions on IoT devices.

Flexible and robust detection for assembly automation with YOLOv5: a case study on HMLV manufacturing line

AbstractAutomating assembly processes in High-Mix, Low Volume (HMLV) manufacturing remains challenging, especially for Small and Medium-sized Enterprises (SMEs). Consequently, many companies still rely on a significant amount of manual operations with an overall low degree of automation. The emergence of artificial intelligence-based algorithms offers potential solutions, enabling assembly automation compatible with multiple products and maintaining overall production flexibility. This paper investigates the application of the YOLO (You Only Look Once) object detection algorithm in an HMLV production line within an SME. The performance of the algorithm was tested for different cases, namely, (a) on different products having similar product features, (b) on completely new products, and (c) under different lighting conditions. The algorithm achieved precision and recall greater than 98$$\%$$ % and mAP50:95 greater than 97$$\%$$ % .

Performance evaluation of a SARS-CoV-2 and influenza A/B combo rapid antigen test

Introduction: The global COVID-19 pandemic and seasonal influenza outbreaks have drawn attention to the critical need for accurate and efficient diagnostic tools.Methods: The performance of the InstaView COVID-19/Flu Ag Combo Test, which was designed to simultaneously detect the SARS-CoV-2, influenza A, and influenza B viruses, was analytically and clinically evaluated.Results: The InstaView COVID-19/Flu Ag Combo Test exhibited robust detection capabilities, accurately identifying SARS-CoV-2, influenza A, and influenza B viruses over a wide concentration range (1.41 × 103 to 7.05 × 104 TCID50/mL). Extensive testing against potential cross-reactants and interferences yielded no false-positive results, indicating the high specificity of the test. Clinical evaluation further confirmed the kit’s reliability, with sensitivity ranging from 95.1% to 98.2% for SARS-CoV-2, 88.9%–95.2% for influenza A, and 91.7%–100% for influenza B depending on the sample type. The specificity was consistently 100% for all of the targeted viruses.Discussion: The InstaView COVID-19/Flu Ag Combo Test thus demonstrated high performance, ease of use, rapid results, and the ability to precisely detect SARS-CoV-2 and influenza A/B infections, making it an effective tool in streamlining diagnostic workflows, optimizing resource allocation, and improving patient outcomes.

Robust quickest change detection in nonstationary processes

Exactly and asymptotically optimal algorithms are developed for robust detection of changes in nonstationary processes. In nonstationary processes, the distribution of the data after change varies with time. The decision maker does not have access to precise information on the post-change distribution. It is shown that if the post-change, nonstationary family has a distribution that is least favorable in a well-defined sense, then the algorithms designed using the least favorable laws are robust optimal. This is the first result in which an exactly robust-optimal solution is obtained in a nonstationary setting where the least favorable law is also allowed to be nonstationary. Examples of nonstationary processes encountered in public health monitoring and space and military applications are provided. Our robust algorithms are also applied to real and simulated data to show their effectiveness.

Development and validation of a generic methyltransferase enzymatic assay based on an SAH riboswitch

Methylation of proteins and nucleic acids plays a fundamental role in epigenetic regulation, and discovery of methyltransferase (MT) inhibitors is an area of intense activity. Because of the diversity of MTs and their products, assay methods that detect S-adenosylhomocysteine (SAH) – the invariant product of S-adenosylmethionine (SAM)-dependent methylation reactions - offer some advantages over methods that detect specific methylation events. However, direct, homogenous detection of SAH requires a reagent capable of discriminating between SAH and SAM, which differ by a single methyl group. Moreover, MTs are slow enzymes and many have submicromolar affinities for SAM; these properties translate to a need for detection of SAH at low nanomolar concentrations in the presence of excess SAM. To meet these needs, we leveraged the exquisite molecular recognition properties of a naturally occurring SAH-sensing RNA aptamer, or riboswitch. By splitting the riboswitch into two fragments, such that SAH binding induces assembly of a trimeric complex, we engineered sensors that transduce binding of SAH into positive fluorescence polarization (FP) and time resolved Förster resonance energy transfer (TR-FRET) signals. The split riboswitch configuration, called the AptaFluor™ SAH Methyltransferase Assay, allows robust detection of SAH (Z’ > 0.7) at concentrations below 10 nM, with overnight signal stability in the presence of typical MT assay components. The AptaFluor assay tolerates diverse MT substrates, including histones, nucleosomes, DNA and RNA, and we demonstrated its utility as a robust, enzymatic assay method for several methyltransferases with SAM Km values < 1 µM. The assay was validated for HTS by performing a pilot screen of 1,280 compounds against the SARS-CoV-2 RNA capping enzyme, nsp14. By enabling direct, homogenous detection of SAH at low nanomolar concentrations, the AptaFluor assay provides a universal platform for screening and profiling MTs at physiologically relevant SAM concentrations.

Implementing robust outlier detection to enhance estimation accuracy of GNSS-IR based seasonal snow depth retrievals

ABSTRACT Monitoring snow depth variations aligned with the seasonal cycle is crucial for studies on climate change and its impacts due to global warming. Especially in terms of snow hydrology, the determination and continuous monitoring of snow depth to determine snow water equivalent is a priority in climate, water science, drought, flood, and inundation prediction studies, particularly in the water cycle. Recently, the so-called Global Navigation Satellite Systems – Interferometric Reflectometry (GNSS-IR) method enables to extract environmental radiometric and geometric characteristics of surface where the signals transmitted from the satellites reflect. The reflected signal follows an extra path than the direct one and causes a major error source for accurate point positioning but a useful tool for sensing the environment. However, outliers remain a challenge in long-term GNSS-IR estimations. In this study, we proposed a median-based robust outlier detection (ROD) approach for identifying outliers in long-term GNSS-IR snow depth estimations. To validate our approach, we analysed 5-year GNSS L1 SNR and L2 SNR data from 1 January 2015 to 31 December 2019 provided by AB33 and AB39 GNSS stations in Alaska, U.S.A. using the GNSS-IR method. We validated GNSS-IR estimations using snow depth measurements from the Coldfoot and Fort Yukon stations in the SNOTEL network. Applying ROD to long-term snow depth estimates increased the highest correlation from 91.58% to 94.93%, and reduced the lowest RMSE from 8.6 cm to 6.7 cm. In addition, the improvement rates calculated to assess the contribution of ROD to the results showed improvements of up to 6.9% in correlation and 26.1% in RMSE. Overall, the results demonstrate that ROD can be effectively used to detect outliers in long-term GNSS-IR snow depth time series.

Human Emotion Detection Using CNN and Transfer Learning

Facial emotion detection is a critical component in human-computer interaction, mental health assessment, and security systems. In this project, we propose a robust facial emotion detection system leveraging state-of-the-art deep learning techniques. Our system utilizes Convolutional Neural Networks (CNNs) to extract meaningful features from facial images and classify them into five distinct emotional categories: neutral, surprise, sad, happy, and angry. We conducted extensive experiments on a diverse dataset consisting of over 10,000 annotated facial images collected from various sources. Through data augmentation techniques such as rotation, translation, and flipping, we expanded the dataset to enhance model training. Additionally, we employed transfer learning by fine-tuning a pre-trained CNN model, ResNet50, on our dataset to leverage its learned features. This project presents a system for real-time emotion monitoring using computer vision techniques. The system utilizes the Haar Cascade Classifier for face detection in live webcam video streams. Furthermore, we evaluated the system's performance across different lighting conditions, poses, and facial occlusions to assess its robustness in real-world scenarios. Our results indicate that the system maintains consistent performance across diverse conditions, making it suitable for deployment in applications requiring real-time facial emotion recognition.

Jellyfish detection algorithm based on multi-gradient flow feature fusion

Jellyfish detection is a challenge in the field of underwater biological identification. Although early works have solved some of the problems of jellyfish detection, there are still shortcomings. Earlier jellyfish datasets were not as extensive, necessitating heavy reliance on data augmentation. Moreover, the detection capability is limited, especially in scenarios involving jellyfish gatherings and occlusions. To overcome these limitations, we establish a high-quality dataset with more jellyfish species and propose a more robust real-time detection algorithm. Our algorithm primarily consists of a multi-gradient flow backbone and a feature fusion module GFPN. Additionally, we have designed a receptive field expansion module SPPFCSPC_G. The entire network employs the FReLU activation function, while the bounding box regression utilizes the WIOU loss function. Our methods demonstrate accuracy and run-time performance in comparison with the state-of-the-art yolo series algorithms. Results show that our algorithm achieves the highest Precision, Recall and mAP50, exceeding the baseline yolov5 by 1.1%, 4.1%, and 4.5%, and outperforming the latest yolov8 by 0.9%, 1.3%, and 2.5%. Importantly, our method effectively addresses aggregation, occlusion, and deformation issues commonly encountered in jellyfish detection.

Heart Disease Prediction and ECG Analysis

Heart disease remains a leading cause of mortality worldwide, necessitating effective detection and management strategies. In this project, we leverage machine learning algorithms to develop a robust heart disease detection system. The dataset used comprises various clinical attributes such as age, gender, chest pain type, and biochemical markers. Through exploratory data analysis and visualization, we gain insights into the dataset's characteristics and correlations between features. Subsequently, we implement several machine learning algorithms, including Logistic Regression, Decision Tree Classifier, K-Nearest Neighbors (KNN), and Support Vector Classifier (SVC), to predict the presence of heart disease based on patient attributes. Model performance is evaluated using metrics such as accuracy score, enabling comparison and selection of the most effective algorithm for heart disease detection. Our findings underscore the potential of machine learning in augmenting traditional diagnostic approaches and paving the way for early intervention and improved patient outcomes in cardiovascular health

Robust detection and refinement of saliency identification

Salient object detection is an increasingly popular topic in the computer vision field, particularly for images with complex backgrounds and diverse object parts. Background information is an essential factor in detecting salient objects. This paper suggests a robust and effective methodology for salient object detection. This method involves two main stages. The first stage is to produce a saliency detection map based on the dense and sparse reconstruction of image regions using a refined background dictionary. The refined background dictionary uses a boundary conductivity measurement to exclude salient object regions near the image's boundary from a background dictionary. In the second stage, the CascadePSP network is integrated to refine and correct the local boundaries of the saliency mask to highlight saliency objects more uniformly. Using six evaluation indexes, experimental outcomes conducted on three datasets show that the proposed approach performs effectively compared to the state-of-the-art methods in salient object detection, particularly in identifying the challenging salient objects located near the image's boundary. These results demonstrate the potential of the proposed framework for various computer vision applications.

Security Information Event Management data acquisition and analysis methods with machine learning principles

In the face of increasing global disruptions, the cybersecurity field is confronting rising threats posed by offensive groups and individual hackers. Traditional security measures often fall short in detecting and mitigating these sophisticated attacks, necessitating advanced intrusion detection methods. The goal of our study is to develop robust network intrusion detection methods using machine learning techniques. In addition, we evaluate the effectiveness of various machine learning models in detecting network intrusions. Model performances are optimized through hyperparameter tuning and feature selection. A range of classification and clustering models have been employed. Data from SIEM systems capturing real-time statistics from cloud-hosted Windows virtual machines has been gathered and augmented with web attack logs from CICIDS2017, each comprising approximately fifteen thousand rows. Hyperparameter tuning, data normalization, standardization and feature selection techniques for model optimization have been used in our study. The research showcases the potential of machine learning in enhancing network intrusion detection capabilities. The findings underscore the effectiveness of the Random Forest Classifier (0.97) and highlight the importance of utilizing diverse datasets and advanced optimization techniques. This study offers valuable insights and sets a foundation for future advancements in cybersecurity strategies and intrusion detection systems.

High-sensitivity liquid chromatography-tandem mass spectrometry quantitative for alkyl imidazolium ionic liquids in human serum: Advancing biomonitoring of human exposure concerns

Alkyl imidazolium ionic liquids (Cn[MIM]), initially heralded as eco-friendly green solvents for diverse industrial applications, have increasingly been recognized fortheir biodegradability challenges and multiple biotoxicity. Despite potential health risks, research into the effects of Cn[MIM] on human health remains scarce, particularly regarding their detection in biological serum samples. This study validated a matrix-matched calibration quantitative method that utilizes solid-phase extraction (SPE) coupled with ultrahigh-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). The method was used to analyze the presence of 10 ionic liquids (ILs) with varying alkyl carbon chain lengths (C2–C12) across 300 human serum samples. Efficient separation was achieved using optimized SPE conditions and a BEH C18 column with an appropriate mobile phase. Results demonstrated a strong linear relationship (0.05–100 ng/mL; R2 = 0.995–0.999), with detection and quantification limits with detection and quantification limits ranging from 0.001 to 0.107 ng/mL and 0.003–0.355 ng/mL, respectively. Intraday and inter-day precisions were 0.85–6.99 % and 1.50–7.46 %, with recoveries between 82 and 113 %. The validated method detected C6MIM in 19 % of samples and C8MIM in 8.3 % of samples, with concentrations ranging from 0.02 to 111.70 μg/L and 0.09–16.99 μg/L, respectively, suggesting a potential risk of human exposure. This underscores the importance of robust detection methods in monitoring environmental and human health impacts of alkyl imidazolium compounds.

Feature Extraction vs Fine-tuning for Cyber Intrusion Detection Model

This study investigates the effectiveness of feature extraction and fine-tuning approaches in developing robust cyber intrusion detection models using the Network-based Security Lab - KDD dataset (NSL-KDD). The role of cyber intrusion detection is pivotal in securing computer networks from unauthorized access and malicious activities. Feature extraction, involving methods such as PCA, LDA, and Autoencoders, aims to transform raw data into informative representations, while fine-tuning leverages pre-trained models for task-specific adaptation. The study follows a comprehensive research method encompassing data collection, preprocessing, model development, and experimental evaluation. Results indicate that LDA and Autoencoders excel in the feature extraction phase, demonstrating precision, high accuracy, F1-Score, and recall. However, fine-tuning a pre-trained Multilayer Perceptron model surpasses individual feature extraction methods, achieving superior performance across all metrics. The discussion emphasizes the complexity and flexibility of these approaches, with fine-tuned models showcasing higher adaptability. In conclusion, this study provides valuable insights into the comparative effectiveness of feature extraction and fine-tuning for cyber intrusion detection. The findings underscore the importance of leveraging pre-trained knowledge and adapting models to specific tasks, offering a foundation for further advancements in enhancing network security through advanced machine learning techniques.

Continuum emission from within the plunging region of black hole discs

ABSTRACT The thermal continuum emission observed from accreting black holes across X-ray bands has the potential to be leveraged as a powerful probe of the mass and spin of the central black hole. The vast majority of existing ‘continuum fitting’ models neglect emission sourced at and within the innermost stable circular orbit (ISCO) of the black hole. Numerical simulations, however, find non-zero emission sourced from these regions. In this work, we extend existing techniques by including the emission sourced from within the plunging region, utilizing new analytical models that reproduce the properties of numerical accretion simulations. We show that in general the neglected intra-ISCO emission produces a hot-and-small quasi-blackbody component, but can also produce a weak power-law tail for more extreme parameter regions. A similar hot-and-small blackbody component has been added in by hand in an ad hoc manner to previous analyses of X-ray binary spectra. We show that the X-ray spectrum of MAXI J1820+070 in a soft-state outburst is extremely well described by a full Kerr black hole disc, while conventional models that neglect intra-ISCO emission are unable to reproduce the data. We believe this represents the first robust detection of intra-ISCO emission in the literature, and allows additional constraints to be placed on the MAXI J1820 + 070 black hole spin which must be low a• &amp;lt; 0.5 to allow a detectable intra-ISCO region. Emission from within the ISCO is the dominant emission component in the MAXI J1820 + 070 spectrum between 6 and 10 keV, highlighting the necessity of including this region. Our continuum fitting model is made publicly available.

A Robust DDoS Intrusion Detection System Using Convolutional Neural Network

In today’s digital age, the proliferation of network-connected devices has triggered a surge in cyberattacks. Distributed Denial-of-Service (DDoS) attacks pose a particularly formidable challenge to network security by disrupting access to vital services. While numerous researchers have proposed DDoS detection methods utilizing machine learning and deep learning techniques, developing a robust and reliable DDoS intrusion detection system remains challenging. This challenge is exacerbated by issues such as highly imbalanced data, multi-classification, and computational complexity. This paper proposes an innovative feature selection approach to create a robust intrusion detection system capable of detecting and classifying recent common DDoS attack types. We evaluate the performance of our model on the CICDDoS2019 benchmark dataset. Our experimental results demonstrate that our proposed model outperforms existing methods, achieving a detection accuracy of 96.82%, a recall of 96.82%, a precision of 96.76%, and an F1 score of 96.50%. Additionally, our model exhibits faster prediction times, with the ability to predict an attack in just 0.189 ms. Notably, our approach, combined with preprocessing and feature selection techniques, outperforms previous works and baseline models in DDoS attack classification.

Benefits of Mann–Kendall trend analysis for vibration-based condition monitoring

Rank-based statistics offer an appealing framework to non-parametric robust trend detection. Especially, the Mann–Kendall test detects the presence of a trend in a series. This paper investigates the benefits of using the Mann–Kendall test, which seems to have remained largely unnoticed in the context of vibration-based condition monitoring. Two contributions structure the present paper. First, theoretical foundations of rank-statistics are reminded and the Z-score is introduced, a statistical metric measuring the presence of a trend in a given series. The performance of the Mann–Kendall test is investigated and confronted to different numerical cases involving series of increasing condition indicators. Second, based on this understanding, the Z-score metric is exploited to design two trend-oriented signal processing tools dedicated to vibration-based condition monitoring. Namely, a new tool named the Mannagram is introduced as a dyadic filterbank representation of the trend of indicators series. It shows effective in the informative band selection problem and unveils more interpretable indicators. Besides, a concise representation of trend by frequency bin is introduced and coined the Kendrum. This representation is shown to help the diagnosis by summarising the trend information of series of spectra. Both methods are demonstrated on run-to-failure series of vibration signals from industrial wind turbines.