No description available.

As the Automotive industry is heavily regulated from a quality point of view, excellence in pro-duction is obligatory. Due to the fact that removing human error from humans is impossible, new solutions must be found. The transition to more data driven production strategies enables the implantation of automated vision systems for replacing humans in simple classification tasks. As research in the field of artificial intelligence advances, the hardware required to run the algorithms decreases. Concurrently small computing platforms break new performance records and the innovation space converges. This work harnesses state-of-the-art from both domains by implementing a plug-on vision system, driven by a resource-constrained edge device in a production line. The implemented CNN-model based on the MobileNetV2 architecture achieved 97.80, 99.93, and 95.67% in accuracy, precision, and recall respectively. The model was trained using only 100 physical samples, which were expanded by a ratio of 1:15 through innovative real world and digital augmentations. The core of the vision system was a commodity device, the Raspberry Pi 4. The solution fulfilled all the requirements while sparking new development ideas for future work.

Improving the energy efficiency of buildings is important for many reasons. There is a large body of data detailing the hourly energy consumption of buildings. This work studies a large data set from the Swedish energy market. This thesis proposes a data analysis methodology for identifying abnormal consumption patterns using two steps of clustering. First, typical weekly energy usage profiles are extracted from each building by clustering week-long segments of the building’s lifetime consumption, and by extracting the medoids of the clusters. Second, all the typical weekly energyusage profiles are clustered using agglomerative hierarchical clustering. Large clusters are assumed to contain normal consumption pattens, and small clusters are assumed to have abnormal patterns. Buildings with a large presence in small clusters are said to be abnormal, and vice versa. The method employs Dynamic Time Warping distance for dissimilarity measure. Using a set of 160 buildings, manually classified by domain experts, this thesis shows that the mean abnormality-score is higher for abnormal buildings compared to normal buildings with p ≈ 0.0036.

Networks connected to the internet are under a constant threat of attacks. To protect against such threats, new techniques utilising already connected hardware have in this thesis been proven to be a viable solution. By equipping network switches with lightweight machine learning models, such as, Decision Tree and Random Forest, no additional devices are needed to be installed on the network.When an attack is detected, the device may notify or take direct actions on the network to protect vulnerable systems. By utilising container software on Westermo's devices, a model has been integrated, limiting its computational resources. Such a system, and its building blocks, are what this thesis has researched and implemented. The system has been validated using multiple different models using a range of parameters.These models have been trained offline on datasets with pre-recorded attacks. The recordings are converted into flows, decreasing dataset size and increasing information density. These flows contain features corresponding to information about the packets and statistics about the flows. During training, a subset of features was selected using a Genetic Algorithm, decreasing the time for processing each packet. After the models have been trained, they are converted to C code, which runs on a network switch. These models are verified online, using a simulated factory, launching different attacks on the network. Results show that the hardware is sufficient for smaller models and that the system is capable of detecting certain types of attacks.

With increasing usage of credit card payments, credit card fraud has also been increasing. Therefore a fast and accurate fraud detection system is vital for the banks. To solve the problem of fraud detection, different machine learning classifiers have been designed and trained on a credit card transaction dataset. However, the dataset is heavily imbalanced which poses a problem for the performance of the algorithms. To resolve this issue, the generative methods Generative Adversarial Network (GAN), Variational Autoencoders (VAE) and Synthetic Minority Oversampling Technique (SMOTE) have been used to generate synthetic samples for the minority class in order to achieve a more balanced dataset. The main purpose of this study is to evaluate the generative methods and investigate the impact of their generated minority samples on the classifiers. The results from this study indicated that GAN does not outperform the other classifiers as the generated samples from VAE were most effective in three out of five classifiers. Also the validation and histogram of the generated samples indicate that the VAE samples have captured the distribution of the data better than SMOTE and GAN. A suggestion to improve on this work is to perform data engineering on the dataset. For instance, using correlation analysis for the features and analysing which features have the greatest impact on the classification and subsequently dropping the less important features and train the generative methods and classifiers with the trimmed down samples. / Med ökande användning av kreditkort som betalningsmetod i världen, har även kreditkort bedrägeri ökat. Därför finns det behov av ett snabbt och tillförligt system för att upptäcka bedrägliga transkationer. För att lösa problemet med att detektera kreditkort bedrägerier, har olika maskininlärnings klassifiseringsmetoder designats och tränats med ett dataset som innehåller kreditkortstransaktioner. Dock är dessa dataset väldigt obalanserade och innehåller mest normala transaktioner, vilket är problematiskt för systemets noggranhet vid klassificering. Därför har generativa metoderna Generative adversarial networks, Variational autoencoder och Synthetic minority oversampling technique använs för att skapa syntetisk data av minoritetsklassen för att balansera datasetet och uppnå bättre noggranhet. Det centrala målet med denna studie var därmed att evaluera dessa generativa metoder och invetigera påverkan av de syntetiska datapunkterna på klassifiseringsmetoderna. Resultatet av denna studie visade att den generativa metoden generative adversarial networks inte överträffade de andra generativa metoderna då syntetisk data från variational autoencoders var mest effektiv i tre av de fem klassifisieringsmetoderna som testades i denna studie. Dessutom visar valideringsmetoden att variational autoencoder lyckades bäst med att lära sig distributionen av orginal datat bättre än de andra generativa metoderna. Ett förslag för vidare utveckling av denna studie är att jobba med data behandling på datasetet innan datasetet används för träning av algoritmerna. Till exempel kan man använda korrelationsanalys för att analysera vilka features i datasetet har störst påverkan på klassificeringen och därmed radera de minst viktiga och sedan träna algortimerna med data som innehåller färre features.

The increasing complexity of trading strategies, when combined with machine learning models, forces market surveillance corporations to develop increasingly sophisticated methods for recognizing potential misuse. One strategy is to employ traders’ weapons against themselves, namely machine learning. However, the data utilized in market surveillance is highly sensitive, what may be available for machine learning is limited. In this thesis, we examine how federated learning for time series data can be used to identify potential market abuse while maintaining client privacy and data security. We are interested in developing a time-series-specific neural network employing federated learning. We demonstrate that when this strategy is used, the performance of detecting potential market abuse is comparable to that of the standard data centralized approach. Specifically, a non-federated model, a federated model, and a federated model with extra data privacy and security protection are evaluated and compared. Each model utilize an LSTM autoencoder to identify market abuse. The results demonstrate that a federated model’s performance in detecting possible market abuse is comparable to that of a non-federated model. Moreover, a federated approach with extra data privacy and security experienced a slight performance loss but is still a competitive model in comparison to the other models. Although this approach results in increased privacy and security, there is a limit to how much privacy and security can be ensured, as excessive privacy led to extremely poor performance. Federated learning offers the ability to increase data privacy and security with little performance decrease. / Den ökande komplexiteten handelsstrategier, i kombination med maskininlärning modeller, tvingar marknadsövervakning företag att utveckla allt mer sofistikerade metoder för att identifiera potentiellt marknadsmissbruk. En strategi är att använda handlarnas vapen mot sig själva, nämligen maskininlärning. Däremot, data som används inom marknadsövervakning är mycket känslig och vad som kan finnas tillgängligt för maskininlärning är begränsat.I den här studien undersöker vi hur federerat lärande för tidsseriedata kan användas till att identifiera potentiellt marknadsmissbruk samtidigt som klienternas integritet och datasäkerhet bibehålls. Vi är intresserade av att utveckla ett tidsserie-specifikt neuralt nätverk med hjälp av federated inlärning. Vi visar att när denna strategi används är prestanda för att upptäcka potentiellt marknadsmissbruk jämförbart med det för den vanliga data-centraliserade metoden. Specifikt, en icke-federerad modell, en federerad modell och en federerad modell med extra dataintegritet och säkerhet utvärderas och jämförs. Varje modell använder en LSTM-Autoencoder för att identifiera marknadsmissbruk. Resultaten visar att en federerad modells prestanda när det gäller att upptäcka eventuellt marknadsmissbruk är jämförbar med en icke-federerad modell. Dessutom, ett federerat tillvägagångssätt med extra dataintegritet upplevde en liten prestandaförlust men är fortfarande en konkurrenskraftig modell i jämförelse med andra modeller. Även om detta tillvägagångssätt resulterar i ökad integritet och säkerhet, finns det en gräns för hur mycket som kan säkerställas. Federated learning möjliggör ökad datasekretess och säkerhet med liten prestandasänkning.

Autonomous vehicles are a rising technology that aims to change the way people think about mobility in the future. A crucial step towards that goal is the assurance that malicious actors cannot instigate accidents that could lead to damages or loss of life. Currently, vehicle platoons, that is vehicles cooperating together to increase fuel saving and driver comfort, are used in limited environments and are the focus of research aimed to make them suitable for real-world wide usage. In that regard, guaranteeing that the vehicle is able to operate alongside other entities, autonomous or not, in the traditional sense is not adequate. The computer systems involved can be the target or the source of a malicious act without the knowledge of the operator in either case. In the context of platooning, these acts can have devastating effects and can originate either from other vehicles on the road or from within, from compromised vehicles that are part of the formation. In this thesis, the focus is centered around the latter. We investigate jamming and data falsification attacks that aim to either destabilize the platoon, thus, reducing its benefits or provoke an accident. These attacks are more difficult to discern and will range from simple falsification attacks to more complex ones that aim to bypass defensive mechanisms. In that sense, we direct our experiments against the platoon maneuvers that are a core functionality of platooning and are required for its nominal operation. The results of this analysis show that several attacks can lead to accidents with position falsification being the most productive. It is also demonstrated that a malicious leader can pose a serious threat to the viability of the platoon because of his unique capability of interacting with all the platoon members. Attacks during the platoon maneuvers are demonstrated to pose a threat, not only to the stability of the formation but also the nature of the platooning application itself. This is achieved by effectively isolating the platoon from potential joiners. / Självkörande fordon är en framväxande teknologi med mål att ändra människors framtida inställning till mobilitet. Ett kritiskt steg mot målet är att försäkra sig om att aktörer med ont uppsåt inte kan orsaka olyckor som kan leda till skador eller dödsfall. För närvarande används fordonståg, alltså fordon som samarbetar för att minska bränsleförbrukning och öka körkomfort, i avgränsade miljöer med fokus på att anpassa dessa för verklig användning. Att garantera att fordonet kan köras tillsammans med andra enheter är då inte tillräckligt eftersom dessa system kan bli mål för externa och interna attacker som kan ha förödande konsekvenser. Denna uppsats fokuserar på det senare fallet och undersöker interna datafalsifierings- och frekvensstörningsattacker avsedda att destabilisera fordonståg i syfte att minska deras fördelar eller provocera fram en olycka. Dessa attacker är svåra att urskilja och inkluderar allt från enkla falsifikationsattacker till komplexa attacker som syftar till att kringgå specifika försvarsmekanismer. Med det i åtanke inriktar vi våra experiment mot de manövrar som är en del av fordonstågens grundfunktionalitet och krävs för deras nominella drift. Resultaten av arbetet visar att under fordonstågmanövrar så kan flertalet av de utvärderade attackerna orsaka olyckor och att attacker genom förfalskning av position var speciellt förödande. Vi har även påvisat att en fordonstågsledare med ont uppsåt utgör ett speciellt allvarligt hot mot fordonstågets funktionalitet på grund av dennes unika möjlighet att interagera med alla medlemmar. Attacker under manövrar har visats utgöra ett hot, inte bara mot stabiliteten av formationen, men även mot de grundläggande egenskaperna hos systemet själv såsom att isolera fordonståget från nya medlemmar.

<p><em>Curvilinear structure detection in images has been investigated for decades. In general, the detection of curvilinear structures includes two aspects, binary segmentation of the image and  inference of the graph representation of the curvilinear network. In our work, we propose a connected-tube model based on a marked point process (MPP) for addressing the two issues. The proposed tube model is applied to fiber detection in microscopy images by combining connected-tube and ellipse models. Moreover, a tube-based segmentation algorithm has been proposed to improve the segmentation accuracy. Experiments on fiber-reinforced polymer images, satellite images, and retinal vessel images will be presented. Additionally, we extend the 2D tube model to a 3D tube model, with each tube be modeled as a cylinder. To investigate the supervised curvilinear structure detection method, we focus on the application of road detection in satellite images and propose a two-stage learning strategy for road segmentation. A probability map is generated in the first stage by a selected neural network, then we attach the probability map image to the original RGB images and feed the resulting four images to a U-Net-like network in the second stage to get a refined result.</em></p> <p><br></p> <p><em>Anomaly detection in time series is a key step in diagnosing abnormal behavior in some systems. Long Short-Term Memory networks (LSTMs) have been demonstrated to be useful for anomaly detection in time series, due to their predictive power. However, for a system with thousands of different time sequences, a single LSTM predictor may not perform well for all the sequences. To enhance adaptability, we propose a stacked predictor framework. Also, we propose a novel dynamic thresholding algorithm based on the prediction errors to extract the potential anomalies. To further improve the accuracy of anomaly detection, we propose a post-detection verification method based on a fast and accurate time series subsequence matching algorithm.</em></p> <p><br></p> <p><em>To detect anomalies from multi-channel time series, a bi-directional transformer-based predictor is applied to generate the prediction error sequences, and a statistical model referred as an anomaly marked point process (Anomaly-MPP) is proposed to extract the anomalies from the error sequences. The effectiveness of our methods is demonstrated by testing on a variety of time series datasets.</em></p>

The increased demand of energy storage systems and electric vehicles on the market result in high demand of lithium-ion batteries. As a lithium-ion battery manufacturer, Northvolt runs quality tests on the products to assess their performance, life and safety. Batteries that are tested are most often behaving as expected, but sometimes deviations occur. Anomaly detection is today most often performed by plotting and comparing produced data to other test-data to find which parameters that are deviating. The purpose of this thesis is to automatize anomaly detection and a proposed solution is to use state-of-the-art machine learning methods. These include using supervised and unsupervised machine learning. Before applying machine learning, the feature engineering is presented. It describes what parameters are extracted from the experiment data sets. Then the supervised machine learning framework is described. For the unsupervised machine learning, a principal component analysis is presented to locate deviations. This thesis also presents a differential capacity analysis, as this could be incorporated with the features in the future. The results shows that the subset of labeled data for supervised learning is too small to produce a model that predicts future deviations. The extracted features are also used in the principal component analysis, where the results show deviations (outliers) and aid targeting the anomalies. These can then be used to determine the root-cause of particular anomalies and mitigate future deviations.

We explore the use of Long short-term memory (LSTM) for anomaly detection in temporal data. Due to the challenges in obtaining labeled anomaly datasets, an unsupervised approach is employed. We train recurrent neural networks (RNNs) with LSTM units to learn the normal time series patterns and predict future values. The resulting prediction errors are modeled to give anomaly scores. We investigate different ways of maintaining LSTM state, and the effect of using a fixed number of time steps on LSTM prediction and detection performance. LSTMs are also compared to feed-forward neural networks with fixed size time windows over inputs. Our experiments, with three real-world datasets, show that while LSTM RNNs are suitable for general purpose time series modeling and anomaly detection, maintaining LSTM state is crucial for getting desired results. Moreover, LSTMs may not be required at all for simple time series. / Vi undersöker Long short-term memory (LSTM) för avvikelsedetektion i tidsseriedata. På grund av svårigheterna i att hitta data med etiketter så har ett oövervakat an-greppssätt använts. Vi tränar rekursiva neuronnät (RNN) med LSTM-noder för att lära modellen det normala tidsseriemönstret och prediktera framtida värden. Vi undersö-ker olika sätt av att behålla LSTM-tillståndet och effekter av att använda ett konstant antal tidssteg på LSTM-prediktionen och avvikelsedetektionsprestandan. LSTM är också jämförda med vanliga neuronnät med fasta tidsfönster över indata. Våra experiment med tre verkliga datasetvisar att även om LSTM RNN är tillämpbara för generell tidsseriemodellering och avvikelsedetektion så är det avgörande att behålla LSTM-tillståndet för att få de önskaderesultaten. Dessutom är det inte nödvändigt att använda LSTM för enkla tidsserier.