Different Types Of Anomalies Research Articles

Adaptive deep learning models for efficient multivariate anomaly detection in IoT infrastructures

Adapting machine learning-based systems to dynamic environments poses significant challenges due to their diverse and rapidly changing nature. Traditional Deep Neural Network (DNN) algorithms often struggle to cope effectively with such variations. This paper presents a novel evolutionary algorithm named Double Evaluation Genetic Evolution (DEGE), specifically tailored to evolve DNNs within dynamic contexts. DEGE represents a pioneering approach in evolutionary computing, focusing on the adaptive evolution of DNN structures across generations. This adaptability plays a crucial role in enabling DNNs to seamlessly adjust to evolving environmental conditions and complexities. To evaluate the efficacy of DEGE, we apply it to the domain of anomaly detection, rigorously testing the adapted DNNs within this specific context. Furthermore, we conduct comparative analyses between DEGE and established optimization methods using standard metrics to elucidate its advantages. Our findings shed light on DEGE’s effectiveness in addressing the challenges posed by dynamic environments, indicating its potential to revolutionize DNN optimization. As a practical application, we integrate DEGE-based DNNs into an IoT anomaly detection system to assess the overall impact of DEGE on anomaly detection performance. Our experiments demonstrate the efficiency of DEGE across 10 generations, showcasing its high adaptability to the dynamism inherent in IoT infrastructures. The proposed DEGE-based anomaly detection system processes highly dynamic environments within IoT infrastructure and classifies/predicts different types of anomalies efficiently with 99% detection accuracy across multiple benchmark and live experiment datasets. Solving the problem of multiclassification in dynamic abnormality detection, the proposed DEGE-based anomaly detection system was highly adaptable to the environment across generations, reaching the optimal DNN structure that delivers the best accuracy, and precision, with minimum loss value.

Communication-efficient federated multi-domain learning for network anomaly detection

Efficient and accurate anomaly detection in a network is of great significance for maintaining network and device security. Most anomaly detection methods assume that different anomalous network data distributions are the same or similar and ignore data privacy preservation. In this paper, a novel Federated Learning (FL) is proposed that it can quickly detect different types of anomalies in Non-Independent and Identically Distributed (Non-IID) data. First, we design a multi-domain machine learning model for multi-domain data, named Aegean, which consists of two modules: an ensemble AutoEncoder (AE) and a Generative Adversarial Network (GAN). Second, because data from different domains are non-IID, we model the anomaly detection problem as a dual problem, which can be recast as a robust optimization problem. The robust optimization problem is non-convex and therefore difficult to solve. As a remedy, we formulate and solve a dual problem by taking the Lagrangian dual function of the original problem. Experiments demonstrate that Aegean significantly outperforms the current state-of-the-art methods, with a 16% F1 score improvement over that of a One-Class Support Vector Machine (OCSVM). The designed FL significantly reduces the communication overhead of FedAvg without sacrificing anomaly detection performance.

Anomaly Detection for Power Quality Analysis Using Smart Metering Systems.

The problem of Power Quality analysis is becoming crucial to ensuring the proper functioning of complex systems and big plants. In this regard, it is essential to rapidly detect anomalies in voltage and current signals to ensure a prompt response and maximize the system's availability with the minimum maintenance cost. In this paper, anomaly detection algorithms based on machine learning, such as One Class Support Vector Machine (OCSVM), Isolation Forest (IF), and Angle-Based Outlier Detection (ABOD), are used as a first tool for rapid and effective clustering of the measured voltage and current signals directly on-line on the sensing unit. If the proposed anomaly detection algorithm detects an anomaly, further investigations using suitable classification algorithms are required. The main advantage of the proposed solution is the ability to rapidly and efficiently detect different types of anomalies with low computational complexity, allowing the implementation of the algorithm directly on the sensor node used for signal acquisition. A suitable experimental platform has been established to evaluate the advantages of the proposed method. All the different models were tested using a consistent set of hyperparameters and an output dataset generated from the principal component analysis technique. The best results achieved included models reaching 100% recall and a 92% F1 score.

Anomalies in the temperature dependences of the time constant of various glass-forming substances

In this work the new precise method for describing the temperature dependences of glass formers which is based on the use of the power law function (JNCS 555 (2021) 120618) is proposed. Analysis of practical temperature dependences using the proposed method made it possible to detect subtle anomalies of fragile relaxation which cannot be detected by other means. Three different types of anomalies were found: cluster-to-cluster (CTC), fragile-to-fragile (FTF), and strong-to-fragile (STF) transitions. The CTC and FTF transitions are described for the first time. The CTC transition is accompanied by changes in activation energy with constant fragility; FTF transition due to changes in fragility and activation energy, as well as the STF transition. The processes occurring in the vicinity of transition points are explained from the point of view of the cluster model.

Congenital Anomalies in American Crocodile (Crocodylus acutus, Cuvier, 1807) Embryos from a Farm Breeder in Colombia.

The American crocodile (Crocodylus acutus, Cuvier, 1807) (Class Reptilia, Family Crocodylidae) is a crocodile species inhabiting the Neotropics. Congenital defects have been described in almost every vertebrate group. In crocodiles, teratology alterations have been described in captive animals (pets, zoos, farms) such as Crocodylus niloticus or Gavialis gangeticus. The present study aimed to characterize congenital malformations of C. acutus from a farm in Lomas de Matunilla, Ballestas, Bolívar, Colombia. A total of 550 unhatched eggs were examined after embryo death. A total of 61 embryos presented malformations, with 42 different types of anomalies observed. Limb and tail malformations (29%) were the most common malformations observed. Several malformations, such as cephalothoracopagus, thoracopagus, sternopagus, xiphopagus twins, campylorrachis scoliosa, and acrania, were documented in crocodiles for the first time. Research in teratology enhances our understanding of crocodile biology. It plays a role in their conservation and management, thus helping to ensure the long-term viability of these species in their natural habitats.

Machine Learning for the Detection and Diagnosis of Anomalies in Applications Driven by Electric Motors.

Manufacturing systems are becoming increasingly flexible, necessitating the adoption of new technologies that allow adaptations to a turbulent and complex modern market. Consequently, modern concepts of production systems require horizontal and vertical integration, extending across value networks and within a factory or production shop. The integration of these environments enables the acquisition of a substantial amount of data containing information pertaining to production, processes, and equipment located on the shop floor. When these data and information are processed and analyzed, they have the potential to reveal valuable insights and knowledge about the manufacturing systems, offering interpretive outcomes for strategic decision making. One of the opportunities presented in this context includes the implementation of predictive maintenance (PdM). However, industrial adoption of PdM is still relatively low. In this paper, the aim is to propose a methodology for selecting the main attributes (variables) to be considered in the instrumentation setup of rotating machines driven by electric motors to decrease the associated costs and the time spent defining them. For this, the most well-known data science and machine learning algorithms are investigated to choose the one most adequate for this task. For the experiments, different testing scenarios were proposed to detect the different possible types of anomalies, such as uncoupled, overloaded, unbalanced, misaligned, and normal. The results obtained show how these algorithms can be effective in classifying the different types of anomalies and that the two models that presented the best accuracy values were k-nearest neighbor and multi-layer perceptron.

Lesion Detection in Optical Coherence Tomography with Transformer-Enhanced Detector.

Optical coherence tomography (OCT) is an emerging imaging tool in healthcare with common applications in ophthalmology for the detection of retinal diseases and in dentistry for the early detection of tooth decay. Speckle noise is ubiquitous in OCT images, which can hinder diagnosis by clinicians. In this paper, a region-based, deep learning framework for the detection of anomalies is proposed for OCT-acquired images. The core of the framework is Transformer-Enhanced Detection (TED), which includes attention gates (AGs) to ensure focus is placed on the foreground while identifying and removing noise artifacts as anomalies. TED was designed to detect the different types of anomalies commonly present in OCT images for diagnostic purposes and thus aid clinical interpretation. Extensive quantitative evaluations were performed to measure the performance of TED against current, widely known, deep learning detection algorithms. Three different datasets were tested: two dental and one CT (hosting scans of lung nodules, livers, etc.). The results showed that the approach verifiably detected tooth decay and numerous lesions across two modalities, achieving superior performance compared to several well-known algorithms. The proposed method improved the accuracy of detection by 16-22% and the Intersection over Union (IOU) by 10% for both dentistry datasets. For the CT dataset, the performance metrics were similarly improved by 9% and 20%, respectively.

Unconventional criticality, scaling breakdown, and diverse universality classes in the Wilson-Cowan model of neural dynamics.

The Wilson-Cowan model constitutes a paradigmatic approach to understanding the collective dynamics of networks of excitatory and inhibitory units. It has been profusely used in the literature to analyze the possible phases of neural networks at a mean-field level, e.g., assuming large fully connected networks. Moreover, its stochastic counterpart allows one to study fluctuation-induced phenomena, such as avalanches. Here we revisit the stochastic Wilson-Cowan model paying special attention to the possible phase transitions between quiescent and active phases. We unveil eight possible types of such transitions, including continuous ones with scaling behavior belonging to known universality classes-such as directed percolation and tricritical directed percolation-as well as six distinct ones. In particular, we show that under some special circumstances, at a so-called "Hopf tricritical directed percolation" transition, rather unconventional behavior is observed, including the emergence of scaling breakdown. Other transitions are discontinuous and show different types of anomalies in scaling and/or exhibit mixed features of continuous and discontinuous transitions. These results broaden our knowledge of the possible types of critical behavior in networks of excitatory and inhibitory units and are, thus, of relevance to understanding avalanche dynamics in actual neuronal recordings. From a more general perspective, these results help extend the theory of nonequilibrium phase transitions into quiescent or absorbing states.

Study of Anomaly Detection in IoT Sensors

Abstract: The rapid proliferation of Internet of Things (IoT) technology has resulted in an exponential increase in the number of connected devices and sensors. These sensors play a crucial role in collecting and transmitting data, enabling various applications and services in diverse domains. However, the large-scale deployment of IoT sensors also introduces new challenges, particularly in the realm of anomaly detection. This research paper presents a comprehensive study of anomaly detection techniques specifically designed for IoT sensors. We delve into the different types of anomalies that can occur in IoT sensor data, including sudden changes, outliers, and malicious attacks. Moreover, we explore the unique characteristics and requirements of IoT sensor networks, such as resource constraints, heterogeneous data, and dynamic network topologies. To address these challenges, we provide an overview of state-of-the-art anomaly detection methods tailored to IoT sensor networks. These methods encompass both traditional statistical approaches and machine learning algorithms, considering their applicability and effectiveness in the IoT context. We discuss the strengths and limitations of each technique, highlighting their suitability for different anomaly detection scenarios. Furthermore, we analyze and compare the performance of these methods using real-world IoT sensor datasets, evaluating their accuracy, efficiency, and scalability. The findings of our study shed light on the strengths and limitations of existing techniques, enabling researchers and practitioners to make informed decisions when choosing an appropriate anomaly detection method for their IoT sensor networks. By enhancing the reliability and security of IoT sensor networks, the outcomes of this research contribute to the advancement of IoT technology and its widespread adoption in various domains, including smart cities, healthcare, transportation, and industrial automation

A machine learning approach for correcting glow curve anomalies in CaSO4:Dy-based TLD dosimeters used in personnel monitoring

The study presents a novel approach to analysing the thermoluminescence (TL) glow curves (GCs) of CaSO4:Dy-based personnel monitoring dosimeters using machine learning (ML). This study demonstrates the qualitative and quantitative impact of different types of anomalies on the TL signal and trains ML algorithms to estimate correction factors (CFs) to account for these anomalies. The results show a good degree of agreement between the predicted and actual CFs, with a coefficient of determination greater than 0.95, a root mean square error less than 0.025, and a mean absolute error less than 0.015. The use of ML algorithms leads to a significant two-fold reduction in the coefficient of variation of TL counts from anomalous GCs. This study proposes a promising approach to address anomalies caused by dosimeter, reader, and handling-related factors. Furthermore, it accounts for non-radiation-induced TL at low dose levels towards improving the dosimetric accuracy in personnel monitoring.

An unsupervised TinyML approach applied to the detection of urban noise anomalies under the smart cities environment

Artificial Intelligence of Things (AIoT) is an emerging area of interest, and this can be used to obtain knowledge and take better decisions in the same Internet of Things (IoT) devices. IoT data are prone to anomalies due to various factors such as malfunctioning of sensors, low-cost devices, etc. Following the AIoT paradigm, this work explores anomaly detection in IoT urban noise sensor networks using a Long Short-Term Memory Autoencoder. Two autoencoder models are trained using normal data from two different sensors in the sensor network and tested for the detection of two different types of anomalies, i.e. point anomalies and collective anomalies. The results in terms of accuracy of the two models are 99.99% and 99.34%. The trained model is quantised, converted to TensorFlow Lite format and deployed on the ESP32 microcontroller (MCU). The inference time on the microcontroller is 4 ms for both models, and the power consumption of the MCU is 0.2693 W ± 0.039 and 0.3268 W ± 0.015. Heap memory consumption during the execution of the program for sensors TA120-T246187 and TA120-T246189 is 528 bytes and 744 bytes respectively.

Detecting Anomalous Graphs in Labeled Multi-Graph Databases

Within a large database 𝒢 containing graphs with labeled nodes and directed, multi-edges; how can we detect the anomalous graphs? Most existing work are designed for plain (unlabeled) and/or simple (unweighted) graphs. We introduce CODEtect , the first approach that addresses the anomaly detection task for graph databases with such complex nature. To this end, it identifies a small representative set 𝒮 of structural patterns (i.e., node-labeled network motifs) that losslessly compress database 𝒢 as concisely as possible. Graphs that do not compress well are flagged as anomalous. CODEtect exhibits two novel building blocks: (i) a motif-based lossless graph encoding scheme, and (ii) fast memory-efficient search algorithms for 𝒮. We show the effectiveness of CODEtect on transaction graph databases from three different corporations and statistically similar synthetic datasets, where existing baselines adjusted for the task fall behind significantly, across different types of anomalies and performance metrics.

GridHTM: Grid-Based Hierarchical Temporal Memory for Anomaly Detection in Videos.

The interest in video anomaly detection systems that can detect different types of anomalies, such as violent behaviours in surveillance videos, has gained traction in recent years. The current approaches employ deep learning to perform anomaly detection in videos, but this approach has multiple problems. For example, deep learning in general has issues with noise, concept drift, explainability, and training data volumes. Additionally, anomaly detection in itself is a complex task and faces challenges such as unknownness, heterogeneity, and class imbalance. Anomaly detection using deep learning is therefore mainly constrained to generative models such as generative adversarial networks and autoencoders due to their unsupervised nature; however, even they suffer from general deep learning issues and are hard to properly train. In this paper, we explore the capabilities of the Hierarchical Temporal Memory (HTM) algorithm to perform anomaly detection in videos, as it has favorable properties such as noise tolerance and online learning which combats concept drift. We introduce a novel version of HTM, named GridHTM, which is a grid-based HTM architecture specifically for anomaly detection in complex videos such as surveillance footage. We have tested GridHTM using the VIRAT video surveillance dataset, and the subsequent evaluation results and online learning capabilities prove the great potential of using our system for real-time unsupervised anomaly detection in complex videos.

Anomaly detection at the European X-ray Free Electron Laser using a parity-space-based method

A novel approach to detect anomalies in superconducting radio-frequency (rf) cavities is presented, based on the parity space method with the goal to detect quenches and distinguish them from other anomalies. The model-based parity space method relies on analytical redundancy and generates a residual signal computed from measurable rf waveforms. The residual is a sensitive indicator of deviation from the model and provides different signatures for different types of anomalies. This new method not only helps with detecting faults but also provides a catalog of unique signatures, based on the detected fault. The method was experimentally verified at the European X-ray Free Electron Laser (EuXFEL). Various types of anomalies incorrectly detected as quenches by the current quench detection system are analyzed using this new approach.9 MoreReceived 28 February 2022Accepted 5 December 2022DOI:https://doi.org/10.1103/PhysRevAccelBeams.26.012801Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.Published by the American Physical SocietyPhysics Subject Headings (PhySH)Research AreasMethods in electromagnetismRadio frequency calculationsAccelerators & Beams

A Unified Bayesian Framework for Joint Estimation and Anomaly Detection in Environmental Sensor Networks

Advanced large-scale environmental monitoring systems relying on the emerging aerial/terrestrial technologies of wireless sensor networks (WSNs), unmanned aerial vehicles (UAVs), and mobile crowdsensing, impose strong requirements on the reliability of the collected data. Unfortunately, sensing units can suddenly suffer unexpected anomalies due to accidental faults or malicious causes. Outlier detection methods have been widely employed to identify and discard unreliable measurements from large data sets, but further improvements in the sensing processes can be obtained by adopting advanced signal processing algorithms that take full advantage of all the collected information without rejecting the measurements. In this paper, we propose a novel unified Bayesian framework that enable simultaneous estimation of a common parameter of interest and identification of multiple and possibly different types of anomalies that can affect sensors in environmental sensor networks. Specifically, we consider two rather general error models based on Gaussian mixtures able to capture different variations affecting the quality of the collected measurements. For each model, we illustrate the optimal joint maximum-likelihood and maximum a-posteriori (ML-MAP) estimation method, which represents the benchmark for the problem at hand, and propose novel reduced-complexity two-step algorithms able to achieve almost the same performance of the joint ML-MAP, but at a fraction of its computational cost. The derivations of all the algorithms are also extended to handle the more general case in which the probability of occurrence of anomalies is unknown and should be inferred from the data using an Empirical Bayes approach. Extensive performance analyses using both synthetic and real experimental data acquired in a network of environmental monitoring stations deployed in the Apulia region, south of Italy, demonstrate the effectiveness of the proposed framework.

Self-Supervised Surface Defect Localization via Joint De-Anomaly Reconstruction and Saliency-Guided Segmentation

Anomaly localization plays one of significant roles in practical industrial applications, but still faces some unique challenges due to the rarity and diversity of anomalies.To address this issue, we propose a self-supervised surface defect localization method via Joint De-anomaly Reconstruction and Saliency-guided Segmentation, also named by JDRSS. Considering the severe lack of anomalous images, an approach for synthesizing anomalous samples is proposed to simulate different types of anomalies, which results in a variety of realistic and diverse anomalous images. To promote the reconstruction quality to the normal reference, we develop a novel auto-encoder based reconstruction network by applying the bundled de-anomaly constraint to refrain both the embedding space and the reconstruction space from the interference of abnormality. Furthermore, in order to accurately localize the surface defect, an anomalous saliency map guided segmentation network is proposed, in which the residuals from the reconstructed image and input are dexterously injected into each segmentation layer as spatial attention, thus enhancing the sensitivity to anomalies. We have conducted extensive experiments on the MVTec anomaly detection dataset and the proposed model achieves considerable performance for different classes of anomaly localization.

AviSense: A Real-time System for Detection, Classification, and Analysis of Aviation Signals

Wireless systems are an integral part of aviation. Apart from their apparent use in air-to-ground communication, wireless systems play a crucial role in avionic functions including navigation and landing. An interference-free wireless environment is therefore critical for the uninterrupted operation and safety of an aircraft. Hence, there is an urgency for airport facilities to acquire the capability to continuously monitor aviation frequency bands for real-time detection of interference and anomalies. To meet this critical need, we design and build AviSense, an SDR-based real-time , versatile system for monitoring aviation bands. AviSense detects and characterizes signal activities to enable practical and effective anomaly detection. We identify and tackle the challenges posed by a diverse set of critical aviation bands and technologies. We evaluate our methodology with real-world aviation signal measurements and two custom datasets of anomalous signals. We find that our signal classification capability achieves a true positive rate of ∼99%, with few exceptions, and a false positive rate of less than 4%. We also demonstrate that AviSense can effectively distinguish between different types of anomalies. We build and evaluate a prototype implementation of AviSense that supports distributed monitoring.

Ontology-based approach to real-time risk management and cyber-situational awareness

The requirement of continuous risk assessment and management is attracting growing attention because of the need of keeping risk under control. Over the years, companies are dealing with a growing number of malicious actions coming from heterogeneous sources, so risk management must be dynamic in real-time to define action strategies and validate the effectiveness of the safeguards in place. This exposure makes it imperative to use sensor-based systems to detect anomalies or to have an updated catalog of vulnerabilities to understand the situation in which the system finds itself and its level of risk. Such a wealth of heterogeneous information has led to the use of ontologies to organize data, as they allow the extraction of new concepts and behaviors, for instance, measuring the risk level of a system or generating metrics for decision support systems. This paper presents an ontology to describe different types of anomalies, merged with previously developed models for Cyber-Threat Intelligence, becoming a proposal to define real-time risk management in a converged secure environment with physical and logical elements, using these ontologies and SPARQL Rules to infer knowledge and calculate dynamically the risk level of the system.

A Comprehensive Study of Local, Global, and Combined Optimization Methods on Synthetic Seismic Refraction and Direct Current Resistivity Data

Most geophysical inversions face the problem of non-uniqueness, which poses a challenge in the mapping and delineation of the subsurface anomalies. To tackle this challenge, a combined local and global optimization approach is considered for jointly inverting two-dimensional direct current resistivity (DCR) and seismic refraction (SR) data that aim to estimate the corresponding physical model parameters. In this combined approach, the output of the local optimization method is used to determine the search space and tuning parameters for the global optimization algorithm. The multi-objective genetic algorithm (non-dominated sorting genetic algorithm) was utilized to jointly optimize the objective functions of two different methods. Because the genetic algorithm is a population-based optimization method, it requires numerous forward calculations. To deal with the expected high computational cost associated with this approach, parallel computing was utilized for the forward function evaluations to reduce the run time of the entire process. The proposed approach was tested using synthetic two-dimensional resistivity and velocity models that had three different types of anomalies (dyke, positive, and combined positive and negative). The results showed an improvement in the anomaly delineation in the output of the combined local and global optimization method compared with the local optimization method. Additionally, similar synthetic models were tested using only the single objective global optimization algorithm (conventional global optimization), which showed promising anomaly delineation.

Detection of atypical response trajectories in biomedical longitudinal databases.

Many health care professionals and institutions manage longitudinal databases, involving follow-ups for different patients over time. Longitudinal data frequently manifest additional complexities such as high variability, correlated measurements and missing data. Mixed effects models have been widely used to overcome these difficulties. This work proposes the use of linear mixed effects models as a tool that allows to search conceptually different types of anomalies in the data simultaneously.