Evaluation of anomaly detection algorithms using machine learning methods

Simple SummaryMachine learning (ML) offers new approaches for analyzing data and is particularly interesting for large datasets. Dairy farmers implement a wide range of sensors, which create large amounts of data, in farming. Therefore, they offer an interesting area for data-driven research. In this review, we show how ML methods have already been used in the scientific literature and describe the potential that these may offer for the future. We found that ML methods were applied to predict data in a variety of areas in dairy farming such as milk yield or energy consumption; however, larger integrated datasets are required to improve the reliability of the algorithms developed.Dairy farmers use herd management systems, behavioral sensors, feeding lists, breeding schedules, and health records to document herd characteristics. Consequently, large amounts of dairy data are becoming available. However, a lack of data integration makes it difficult for farmers to analyze the data on their dairy farm, which indicates that these data are currently not being used to their full potential. Hence, multiple issues in dairy farming such as low longevity, poor performance, and health issues remain. We aimed to evaluate whether machine learning (ML) methods can solve some of these existing issues in dairy farming. This review summarizes peer-reviewed ML papers published in the dairy sector between 2015 and 2020. Ultimately, 97 papers from the subdomains of management, physiology, reproduction, behavior analysis, and feeding were considered in this review. The results confirm that ML algorithms have become common tools in most areas of dairy research, particularly to predict data. Despite the quantity of research available, most tested algorithms have not performed sufficiently for a reliable implementation in practice. This may be due to poor training data. The availability of data resources from multiple farms covering longer periods would be useful to improve prediction accuracies. In conclusion, ML is a promising tool in dairy research, which could be used to develop and improve decision support for farmers. As the cow is a multifactorial system, ML algorithms could analyze integrated data sources that describe and ultimately allow managing cows according to all relevant influencing factors. However, both the integration of multiple data sources and the obtainability of public data currently remain challenging.

Internet of Medical Things (IoMT) envisions the interconnection of healthcare devices and sensors to enable a new class of applications relying on medical data processing and storage. Moreover, machine learning and deep learning algorithms applied in healthcare domain allow health professionals to monitor, diagnose, focus and highlight the region of the problem and propose the required and accurate solution in the shortest duration possible. This paper surveys the state of art machine and deep learning algorithms in IoT healthcare.

Artificial intelligence (AI) has transformed the world and influences every aspect of our lives. AI has revolutionized industries, healthcare, and daily life tasks. Machine learning (ML) algorithms are used in AI to mimic how people think, reason, and learn. ML techniques can be classified into four categories based on the learning method and the desired outcome: supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. A subfield of ML called deep learning (DL). A DL algorithm uses layers of neural networks to learn complex patterns from large amounts of data through advances in computing power and improved training techniques. In the healthcare sector, AI is an excellent tool for supporting the treatment of the growing population by decreasing the pressure on healthcare providers. In the medical field, diagnostic radiology is an area most in need of AI. Diagnostic imaging applications utilizing AI are promising, but they are constrained by a number of limitations. Data availability, for example, is one of the biggest challenges in programming AI algorithms. In the United Arab Emirates (UAE), AI in radiology has revolutionized medical imaging by enhancing diagnostic accuracy and streamlining workflows. It enables early disease detection and empowers radiologists to offer personalized treatment plans. However, it is worth noting that while AI adoption is on the rise, its full potential has not yet been fully realized. In this chapter, we will discuss types of machine learning algorithms, applications in healthcare focusing on diagnostic radiology, and their limitations.

The increasing digitalization across all business sectors creates ever larger amounts of data. When analyzing this data to extract information, traditional data analysis methods easily meet their limits of performance. Conversely, methods from the machine learning (ML) spectrum promise to be a versatile tool for solving highly complex data-related tasks. Yet, organizations fail to identify relevant applications for ML due to a lack of systematic understanding of the applicability of the technology. This research paper draws upon the assumption that real-world problems exhibit distinctive characteristics that indicate their suitability for the application of ML. A framework for describing those constituent problem characteristics is proposed. Investigating the functional differences between traditional data analysis methods and ML, constituent characteristics are derived based on the distinctive technological abilities of ML. In order to differentiate simple ML methods from advanced ML methods regarding their technological abilities, a framework is presented. We suggest that advanced ML methods such as Deep Learning or Transfer Learning provide different potential benefits than simple methods such as Decision Trees, therefore necessitating this additional distinction.

Chemical heuristics have been fundamental to the advancement of chemistry and materials science. These heuristics are typically established by scientists using knowledge and creativity to extract patterns from limited datasets. Machine learning offers opportunities to perfect this approach using computers and larger datasets. Here, we discuss the relationships between traditional heuristics and machine learning approaches. We show how traditional rules can be challenged by large-scale statistical assessment and how traditional concepts commonly used as features are feeding the machine learning techniques. We stress the waste involved in relearning chemical rules and the challenges in terms of data size requirements for purely data-driven approaches. Our view is that heuristic and machine learning approaches are at their best when they work together.

The subject of the article is the study and comparison of two approaches to modelling the battery discharge of a CubeSat satellite: analytical using equivalent circuit and machine learning. The article aims to make a reasoned choice of the approach to modelling the battery discharge of a CubeSat satellite. Modelling the battery discharge of a satellite will enable the prediction of the consequences of disconnecting the autonomous power system and ensure the fault tolerance of equipment in orbit. Therefore, the selected study is relevant and promising. This study focuses on the analysis of CubeSat satellite data, based explicitly on orbital data samples of the power system, which include data available at the time of the article’s publication. The dataset contains data on the voltage (mV), current (mA), and temperature (Celsius) of the battery and solar panels attached to the five sides of the satellite. In this context, two approaches are considered: analytical modelling based on physical laws and machine learning, which uses empirical data to create a predictive model. Results: A comparative analysis of the modeling results reveals that the equivalent circuit approach has the advantage of transparency, as it identifies possible parameters that facilitate understanding of the relationships. However, the model is less flexible to environmental changes or non-standard satellite behavior. The machine learning model demonstrated more accurate results, as it can account for complex dependencies and adapt to actual conditions, even when they deviate from theoretical assumptions. However, the model requires prior training on a large amount of data and is less well understood in terms of physical laws. General conclusions. The equivalent circuit approach provides high accuracy and reliability under known conditions, but it is limited when external parameters change. The machine learning approach demonstrates better overall accuracy and stability, especially under variable or unpredictable conditions, but requires a large amount of high-quality data and more complex interpretation. Thus, the most effective approach may be a hybrid one, where the analytical model serves as the basis and machine learning is used as a tool for refining or compensating for inaccuracies.

In recent years, infections and damage caused by malware have increased at exponential rates. At the same time, machine learning (ML) techniques have shown tremendous promise in many domains, often out performing human efforts by learning from large amounts of data. Results in the open literature suggest that ML is able to provide similar results for malware detection, achieving greater than 99% classification accuracy [49]. However, the same detection rates when applied in deployed settings have not been achieved. Malware is distinct from many other domains in which ML has shown success in that (1) it purposefully tries to hide, leading to noisy labels and (2) often its behavior is similar to benign software only differing in intent, among other complicating factors. This report details the reasons for the difficultly of detecting novel malware by ML methods and offers solutions to improve the detection of novel malware. We propose to detect malware by detecting behaviors commonly exhibited by malware such as DLL injection, and process hollowing. This is based on the assumption that there is a set of behaviors that are common to most malware samples and detecting them will generalize to novel malware. Additionally, detected behaviors point analysts toward appropriate handling and mitigation strategies, which is not the case with a binary benign/malicious classification. A behavior labeling method was developed and was used to label an existing malware dataset. Results show that detecting malicious behaviors is much more difficult than simply classifying malware and goodware?achieving 80% accuracy compared to reported 99% accuracy from classifying malware and goodware. This drop is due to several reasons which are detailed in the report. We also propose to evaluate the performance of detecting novel malware by holding out a malware family for testing and training on the other families. Traditional ML evaluation will shuffle the data and then split the data into training and testing. Our method addresses the use-case when novel malware families are encountered and they require more than just a malicious or benign designation. Our results suggest that this type of evaluation is much more difficult than traditional methods and provides more realistic results, albeit, significantly worse. For our behavior detection, accuracy decreases from 80% to 68% across all behaviors when holding out a malware family from training. We show that the degradation in performance is because each malware family has distinct characteristics resulting in high extrapolations by an ML model. Here, an ML model should return an "I do not know" response and request further analysis from an analyst. We run a number of experiments that compare novel malware families to the training data using different feature representations including a genomics-inspired distance measure and features extracted by deep learning. Generally, held-out families are significantly different from the training data, resulting in unpredictable results. This has been observed generally in the ML community [22, 9]. We empirically demonstrate this in the domain of malware detection. In an attempt to improve the detection of malware behaviors, we examine the impact that additional synthetic data has on the performance of an ML model in detecting behaviors in novel malware families. We find that while synthetic data does improve the performance of ML models, often simpler methods perform better than more complicated ones. Two generative modeling techniques were examined to produce synthetic malware samples such that the behaviors present are able to be specified externally. The difficulty is due to finer grained analysis of the executable and modifying the problem from a binary classification problem to a multi-label problem. The addition of synthetic data increases the overall accuracy from 68% to 70%. While far less accurate than measures presented in academic analyses, we believe that this is more representative of real-world performance and allows models to be properly placed within a malware detection system. We suggest that in highly dynamic environments ML pipelines should determine whether an ML model is competent in the area of new data and should involve mechanisms to improve over time with a human in the loop.

The Internet of Things is rapidly growing with the demand for low-power, long-range wireless communication technologies. Long Range Wide Area Network (LoRaWAN) is one such technology that has gained significant attention in recent years due to its ability to provide long-range communication with low power consumption. One of the main issues in LoRaWAN is the efficient utilization of radio resources (e.g., spreading factor and transmission power) by the end devices. To solve the resource allocation issue, machine learning (ML) methods have been used to improve the LoRaWAN network performance. The primary aim of this survey paper is to study and examine the issue of resource management in LoRaWAN that has been resolved through state-of-the-art ML methods. Further, this survey presents the publicly available LoRaWAN frameworks that could be utilized for dataset collection, discusses the required features for efficient resource management with suggested ML methods, and highlights the existing publicly available datasets. The survey also explores and evaluates the Network Simulator-3-based ML frameworks that can be leveraged for efficient resource management. Finally, future recommendations regarding the applicability of the ML applications for resource management in LoRaWAN are illustrated, providing a comprehensive guide for researchers and practitioners interested in applying ML to improve the performance of the LoRaWAN network.

The modern developments in the Internet of Things (IoT) have been producing an increasing number of interconnected devices, allowing a variety of elegant applications. These IoT devices generate huge information that require smart data study and processing methods, for instance, deep learning (DL). Remarkably, the DL algorithms, when amalgamated with the industrial IoT (IIoT), can facilitate a mixture of applications, for example, elegant assembling, effective manufacturing, efficient networking, accident identifying, and prevention of errors. Also, IIoT can change the way in which industries use to produce autonomous self-healing machines and improve stocks using machine learning (ML). Alternatively, ML carries hidden imminence in IoT data for quick, automated answers, and enhanced assessment making. ML for IoT can be used to venture future trends, become aware of anomalies, and augment intelligence by consuming images, videos, and audios. Various industries currently use ML in IoT to keep up acceptable standards as well as maintain predictable things in terms of benefits and profits at different times. Banks and other businesses in the financial industry make use of ML technology for two main reasons: to recognise significant imminence in data and to avoid scams. Data mining can also be used to make out clients with high-risk outlines or use cyber-surveillance to locate notifying indications of fraud. IIoT has been a means to digital transformation in manufacturing. It takes up a network of sensors to implement a serious building of data and brings cloud software into action to use this data in expensive approaches about the effectiveness of manufacturing procedures. Machines can gain knowledge of the data and algorithms that are accountable for origin faults in the system to identify the errors before those can do damage. Manufacturers can make use of ML to get better safeguarding processes and facilitate them to construct real-time, smart judgments derived from data. So the main aim of this chapter is to focus on data acquisition, integration, and predictive control in the industry using ML for IIoTs for real-time collection and analysis of thousands of measurement points, an early detection of equipment failures and process disturbances, detection of new types of anomalies as they occur for the first time, and Web-based UI for handling alerts.

Today’s Machine Learning (ML) in a blend with Internet of Things (IoT)-based cloud applications plays a significant role in our everyday life. As indicated by Gartner’s recent study, there are around 25 billion devices and a gadget interfacing with IoT including wearables and automated vehicles to smart homes and smart cities applications. All such connected (smart) devices generate immense data that needs to be examined and analysed, to ensure that they continually learn from the available datasets and better themselves without any manual interference. This is where the prerequisite for machine learning comes into being. Several ML algorithms and techniques are introduced in a short time to easily evaluate big data measurements, increasing the IoT’s productivity. Similarly, special ML techniques, such as decision trees, clustering and neural and Bayesian networks, allow devices and gadgets to discern trends from various sources in different kinds of datasets and take appropriate decisions based on their analysis.It would really be difficult for smart devices to make smart decisions gradually without including and enforcing ML. The IoT helps to interconnect various hardware devices, such as houses, cars, electronic gadgets and other devices that are integrated with actuators, sensors and software, so that data can be collected and shared. As various organizations understand the progressive capability of the IoT, they have begun finding various obstructions they have to deliver to use it productively. Numerous organizations and businesses use ML to exploit the IoT’s latent capacity. This chapter evaluates the different methods of machine learning that deal with the challenges posed in the handling of IoT data. Note that this big data is generated through the communication of Internet of Things/smart devices, and this data stored at cloud. The taxonomy of machine learning algorithms is described in this chapter, explaining how different techniques are applied to data generated using IoT devices. It will also address the future problems of machine learning for IoT data analytics.KeywordsInternet of Things (IoT)-based cloud applicationsMachine learning techniquesArtificial intelligenceClassificationSmart era

A defect detection system for wire arc additive manufacturing using incremental learning

The Internet is a primary breeding ground for various network attacks, presenting a persistent challenge to cybersecurity. Machine learning has emerged as a powerful tool for automating the detection of cyber attacks and threats. Unlike traditional algorithms that rely on predefined signatures or rules, machine learning excels at uncovering complex patterns within vast volumes of Internet data. This capability enables the development of more intelligent and effective security solutions. The integration of machine learning into security systems has become a critical focus in both academia and industry. Nevertheless, the integration of machine learning and cybersecurity remains in its infancy. A significant gap exists between theory and practice: while machine learning may perform well in controlled or simulated environments, it may not work as expected in complex and dynamic real-world scenarios. Currently, machine learning security solutions are often evaluated using publicly available datasets in controlled laboratory settings. These lab-based evaluations introduce uncertainty about their practical effectiveness, which may hinder their adoption by security practitioners in real-world contexts. Proactive security measures aim to prevent threats before they cause harm, offering greater efficiency and effectiveness compared to reactive security, which relies on responsive strategies. Machine learning can enhance proactive security by automatically identifying potential vulnerabilities and predicting emerging threats, thereby strengthening the protection of Internet infrastructures. This dissertation aims to facilitate the application of machine learning in real-world cybersecurity settings, with a focus on addressing five core research questions: 1) What general challenges and limitations arise in the design, development, and deployment of practical ML-based security applications? 2) How can ML-based security solutions be effectively implemented in real-world scenarios? 3) What specific challenges and limitations are encountered when implementing such solutions in practical environments? 4) What current technologies, approaches, or solutions can be used to address these challenges and limitations, thereby facilitating the implementation of practical ML-based security systems? 5) What future research directions hold promise for advancing the practical application of machine learning in the cybersecurity field? We begin with a comprehensive literature review, identifying potential challenges and limitations in the practical application of machine learning (ML) within the field of cybersecurity. Building on these insights, we develop two practical tools designed to enhance existing proactive security mechanisms, directly addressing specific real-world challenges. Subsequently, we offer key recommendations regarding methodologies and solutions to mitigate obstacles in the development and deployment of realistic ML-based security systems. Finally, we propose future research directions, highlighting promising technologies and concepts with the potential to advance the field of practical ML applications in cybersecurity. This dissertation represents the first comprehensive investigation into the application of machine learning in realistic cybersecurity scenarios, with significant implications for the field of ML-based cybersecurity. It provides security researchers and practitioners with a deeper understanding of practical challenges and obstacles in this field. The development of two practical tools not only advances state-of-the-art proactive security mechanisms but also delivers actionable recommendations to assist researchers in addressing these challenges. Finally, the proposed future research directions highlight the potential of emerging technologies to enable more effective ML-driven security solutions in real-world environments. Overall, the insights presented in this dissertation help bridge the gap between theory and practice in ML-based cybersecurity, promoting the practical implementation of ML security systems that can replace or complement traditional approaches.

In the era of artificial intelligence, machine learning methods are successfully used in various fields. Machine learning has attracted extensive attention from investors in the financial market, especially in stock price prediction. However, one argument for the machine learning methods used in stock price prediction is that they are black-box models which are difficult to interpret. In this paper, we focus on the future stock price prediction with the historical stock price by machine learning and deep learning methods, such as support vector machine (SVM), random forest (RF), Bayesian classifier (BC), decision tree (DT), multilayer perceptron (MLP), convolutional neural network (CNN), bi-directional long-short term memory (BiLSTM), the embedded CNN, and the embedded BiLSTM. Firstly, we manually design several financial time series where the future price correlates with the historical stock prices in pre-designed modes, namely the curve-shape-feature (CSF) and the non-curve-shape-feature (NCSF) modes. In the CSF mode, the future prices can be extracted from the curve shapes of the historical stock prices. Conversely, in the NCSF mode, they can’t. Secondly, we apply various algorithms to those pre-designed and real financial time series. We find that the existing machine learning and deep learning algorithms fail in stock price prediction because in the real financial time series, less information of future prices is contained in the CSF mode, and perhaps more information is contained in the NCSF. Various machine learning and deep learning algorithms are good at handling the CSF in historical data, which are successfully applied in image recognition and natural language processing. However, they are inappropriate for stock price prediction on account of the NCSF. Therefore, accurate stock price prediction is the key to successful investment, and new machine learning algorithms handling the NCSF series are needed.

This study presents a comparative analysis of the classification performance of facial and emotion recognition systems using Machine Learning (ML) and Deep Learning (DL) algorithms. The primary objective of this work is to evaluate the applicability of emotion recognition in fields such as psychotherapy and crime analysis, using the FER-2013 dataset.The study was conducted with ML algorithms such as Support Vector Machines (SVM), Random Forest (RF), K-Nearest Neighbor (KNN), Decision Trees, and Gradient Boosting, as well as the DL algorithm, Convolutional Neural Network (CNN). Supported by the application of feature selection, preprocessing, and feature extraction techniques, the model’s performance was measured using standard metrics such as accuracy, precision, F1-score, and AUC-ROC.The experimental results showed that the highest classification accuracy (46.61%) was achieved with the CNN model. While the ML models generally offered lower accuracy, they provided advantages in terms of computational efficiency in specific scenarios.This study aims to contribute to the literature by providing a comparative analysis of ML and DL algorithms and by highlighting the effect of data preprocessing on performance. The findings set targets for future work, such as real-time system integration and hybrid model development.

Network information security risks multiply and become more dangerous. Hackers today generally target end-to-end technology and take advantage of human weaknesses. Furthermore, hackers take advantage of technology weaknesses by applying various methods to attack. Nowadays, one of the greatest dangers to the modern digital world is malicious URLs, and stopping them is one of the biggest challenges in the field of cyber security. Detecting harmful URLs using machine learning and deep learning algorithms have been the subject of various academic papers. However, time and accuracy are the two biggest challenges of these tools. This paper proposes a multilayer perceptron (MLP) model that utilizes two significant aspects to make it more practical, lightweight, and fast: Using only lexical features and a decision tree (DT) algorithm to select the best relevant subset of features. The effectiveness of the experimental outcomes is evaluated in terms of time, accuracy, and error reduction. The results show that a MLP model using 35 features could achieve an accuracy of 94.51% utilizing only URL lexical features. Furthermore, the model is improved in time after applying the DT as feature selection with a slight improvement in accuracy and loss.