The aim of the project was to examine the feasibilityof evading continuous authentication systems with a generativeadversarial network. To this end, a group of supervised andunsupervised state-of-the-art classifiers were trained on a publiclyavailable dataset of stroke patterns on mobile devices. To find thebest configurations for each classifier, hyper-parameter searcheswere performed. To attack the classifiers, a generative adversarialnetwork was trained on the dataset to reproduce samples followingthe same distribution. The generative adversarial networkwas optimized to maximize the Equal Error Rate metric of theclassifiers on the reproduced data. Our results show that theEqual Error Rate and the Threshold False Acceptance Rateincreased on generated samples compared to random evasionattacks. Across the classifiers, the greatest increase in Equal ErrorRate was 26 percent (for the artificial neural network), and thegreatest increase in Threshold False Acceptance Rate was 60percent for the same classifier. Moreover, it was found that, ingeneral, the unsupervised classifiers were more robust towardsthis type of attack. The results indicate that evasion attacksagainst continuous authentication systems using a generativeadversarial network are feasible and thus constitute a real threat. / Målet med detta projekt var att undersökamöjligheten att undgå ett aktivt verifieringssystem med hjälpav ett generativt nätverk. För att göra detta valde vi ut ettantal moderna klassifieringsalgoritmer och tränade dem på enoffentlig datasamling av svepmönster på mobiltelefoner. För atterhålla de bästa konfigurationerna för varje klassifieringsalgoritmutfördes hyper-parameter sökningar. För att attackera klassifieringsalgorithmernaimplementerades ett generative adversarialnetwork som tränades på datasamlingen för att reproduceraliknande svepmönster. Det generativa nätverket optimerades föratt maximera klassifieringsalgoritmernas likvärdiga felkvot medden reproducerade datan. Resultaten visar att den likvärdigafelkvoten och tröskeln av den felaktiga verifieringskvoten ökademed den reproducerade datan jämfört med slumpmässiga tester.Den högsta ökningen av den likvärdiga felkvoten var 26 procent(för det artificiella neurala nätverket) och den högsta ökningenav tröskeln av den felaktiga verifieringskvoten var 60 procent forsamma algoritm. Därutöver fann vi att de oövervakade klassifieringsalgoritmernavar mer motståndskraftiga mot denna typenav attack jämfört med de övervakade klassifieringsalgoritmerna.Resultaten tyder på att det är möjligt att till viss del undgå ettaktivt verifieringssystem med hjälp av ett generative adversarialnetwork och att denna typen av attacker utgör ett konkret hot. / Kandidatexjobb i elektroteknik 2021, KTH, Stockholm

An accurate measurement of bubbly flow has a significant value for understanding the bubble behavior, heat and energy transfer pattern in different engineering systems. It also helps to advance the theoretical model development in two-phase flow study. Due to the interaction between the gas and liquid phase, the flow patterns are complicated in recorded image data. The segmentation and reconstruction of overlapping bubbles in these images is a challenging task. This dissertation provides a complete set of image processing algorithms for bubbly flow measurement. The developed algorithm can deal with bubble overlapping issues and reconstruct bubble outline in 2D high speed images under a wide void fraction range. Key bubbly flow parameters such as void fraction, interfacial area concentration, bubble number density and velocity can be computed automatically after bubble segmentation. The time-averaged bubbly flow distributions are generated based on the extracted parameters for flow characteristic study. A 3D imaging system is developed for 3D bubble reconstruction. The proposed 3D reconstruction algorithm can restore the bubble shape in a time sequence for accurate flow visualization with minimum assumptions. The 3D reconstruction algorithm shows an error of less than 2% in volume measurement compared to the syringe reading. Finally, a new image synthesis framework called Bubble Generative Adversarial Networks (BubGAN) is proposed by combining the conventional image processing algorithm and deep learning technique. This framework aims to provide a generic benchmark tool for assessing the performance of the existed image processing algorithms with significant quality improvement in synthetic bubbly flow image generation. / Ph. D. / Bubbly flow phenomenon exists in a wide variety of systems, for example, nuclear reactor, heat exchanger, chemical bubble column and biological system. The accurate measurement of the bubble distribution can be helpful to understand the behaviors of these systems. Due to the complexity of the bubbly flow images, it is not practical to manually process and label these data for analysis. This dissertation developed a complete suite of image processing algorithms to process bubbly flow images. The proposed algorithms have the capability of segmenting 2D dense bubble images and reconstructing 3D bubble shape in coordinate with multiple camera systems. The bubbly flow patterns and characteristics are analyzed in this dissertation. Finally, a generic image processing benchmark tool called Bubble Generative Adversarial Networks (BubGAN) is proposed by combining the conventional image processing and deep learning techniques together. The BubGAN framework aims to bridge the gap between real bubbly images and synthetic images used for algorithm benchmark and algorithm.

Autonomous vehicles rely on sensors for a clear understanding of the environment and in a heavy duty truck, the sensors are placed at multiple locations like the cabin, chassis and the trailer in order to increase the field of view and reduce the blind spot area. Usually, these sensors perform best when they are stationary relative to the ground, hence large and fast movements, which are quite common in a truck, may lead to performance reduction, erroneous data or in the worst case, a sensor failure. This enforces a need to validate the sensors before using them for making life-critical decisions. This thesis proposes Domain Adaptation as one of the strategies to co-validate Inertial Measurement Unit (IMU) sensors. The proposed Generative Adversarial Network (GAN) based framework predicts the data of one IMU using other IMUs in the truck by implicitly learning the internal dynamics. This prediction model along with other sensor fusion strategies would be used by the supervising system to validate the IMUs in real-time. Through data collected from real-world experiments, it is shown that the proposed framework is able to accurately transform raw IMU sequences across domains. A further comparison is made between Long Short Term Memory (LSTM) and WaveNet based architectures to show the superiority of WaveNets in terms of performance and computational efficiency. / Autonoma fordon förlitar sig på sensorer för att skapa en bild av omgivningen. På en tung lastbil placeras sensorerna på multipla ställen, till exempel på hytten, chassiet och på trailern för att öka siktfältet och för att minska blinda områden. Vanligtvis presterar sensorerna som bäst när de är stationära i förhållande till marken, därför kan stora och snabba rörelser, som är vanliga på en lastbil, leda till nedsatt prestanda, felaktig data och i värsta fall fallerande sensorer. På grund av detta så finns det ett stort behov av att validera sensordata innan det används för kritiskt beslutsfattande. Den här avhandlingen föreslår domänadaption som en av de strategier för att samvalidera Tröghetsmätningssensorer (IMU-sensorer). Det föreslagna Generative Adversarial Network (GAN) baserade ramverket förutspår en Tröghetssensors data genom att implicit lära sig den interna dynamiken från andra Tröghetssensorer som är monterade på lastbilen. Den här prediktionsmodellen kombinerat med andra sensorfusionsstrategier kan användas av kontrollsystemet för att i realtid validera Tröghetssensorerna. Med hjälp av data insamlat från verkliga experiment visas det att det föreslagna ramverket klarar av att med hög noggrannhet konvertera obehandlade Tröghetssensor-sekvenser mellan domäner. Ytterligare en undersökning mellan Long Short Term Memory (LSTM) och WaveNet-baserade arkitekturer görs för att visa överlägsenheten i WaveNets när det gäller prestanda och beräkningseffektivitet.

Generative adversarial networks (GANs), a type of generative modeling framework, has received much attention in the past few years since they were discovered for their capacity to recover complex high-dimensional data distributions. These provide a compressed representation of the data where all but the essential features of a sample is extracted, subsequently inducing a similarity measure on the space of data. This similarity measure gives rise to the possibility of interpolating in the data which has been done successfully in the past. Herein we propose a new stochastic interpolation method for GANs where the interpolation is forced to adhere to the data distribution by implementing a sequential Monte Carlo algorithm for data sampling. The results show that the new method outperforms previously known interpolation methods for the data set LINES; compared to the results of other interpolation methods there was a significant improvement measured through quantitative and qualitative evaluations. The developed interpolation method has met its expectations and shown promise, however it needs to be tested on a more complex data set in order to verify that it also scales well. / Generative adversarial networks (GANs) är ett slags generativ modell som har fått mycket uppmärksamhet de senaste åren sedan de upptäcktes för sin potential att återskapa komplexa högdimensionella datafördelningar. Dessa förser en komprimerad representation av datan där enbart de karaktäriserande egenskaperna är bevarade, vilket följdaktligen inducerar ett avståndsmått på datarummet. Detta avståndsmått möjliggör interpolering inom datan vilket har åstadkommits med framgång tidigare. Häri föreslår vi en ny stokastisk interpoleringsmetod för GANs där interpolationen tvingas följa datafördelningen genom att implementera en sekventiell Monte Carlo algoritm för dragning av datapunkter. Resultaten för studien visar att metoden ger bättre interpolationer för datamängden LINES som användes; jämfört med resultaten av tidigare kända interpolationsmetoder syntes en märkbar förbättring genom kvalitativa och kvantitativa utvärderingar. Den framtagna interpolationsmetoden har alltså mött förväntningarna och är lovande, emellertid fordras att den testas på en mer komplex datamängd för att bekräfta att den fungerar väl även under mer generella förhållanden.

This study investigates synthetic schematic generation using conditional generative adversarial networks, specifically the Pix2Pix algorithm was implemented for the experimental phase of the study. With the increase in deep neural network’s capabilities and availability, there is a demand for verbose datasets. This in combination with increased privacy concerns, has led to synthetic data generation utilization. Analysis of synthetic images was completed using a survey. Blueprint images were generated and were successful in passing as genuine images with an accuracy of 40%. This study confirms the ability of generative neural networks ability to produce synthetic blueprint images.

Speech synthesis is a technology that increasingly influences our daily lives, in the form of smart assistants, advanced translation systems and similar applications. In this thesis, the phenomenon of making one’s voice sound like the voice of someone else is explored. This topic is called voice conversion and needs to be done without altering the linguistic content of speech. More specifically, a Cycle-Consistent Adversarial Network that has proven to work well in a monolingual setting, is evaluated in a multilingual environment. The model is trained to convert voices between native speakers from the Nordic countries. In the experiments no parallel, transcribed or aligned speech data is being used, forcing the model to focus on the raw audio signal. The goal of the thesis is to evaluate if performance is degraded in a multilingual environment, in comparison to monolingual voice conversion, and to measure the impact of the potential performance drop. In the study, performance is measured in terms of naturalness and speaker similarity between the generated speech and the target voice. For evaluation, listening tests are conducted, as well as objective comparisons of the synthesized speech. The results show that voice conversion between a Swedish and Norwegian speaker is possible and also that it can be performed without performance degradation in comparison to Swedish-to-Swedish conversion. Furthermore, conversion between Finnish and Swedish speakers, as well as Danish and Swedish speakers show a performance drop for the generated speech. However, despite the performance decrease, the model produces fluent and clearly articulated converted speech in all experiments. These results are noteworthy, especially since the network is trained on less than 15 minutes of nonparallel speaker data for each speaker. This thesis opens up for further areas of research, for instance investigating more languages, more recent Generative Adversarial Network architectures and devoting more resources to tweaking the hyperparameters to further optimize the model for multilingual voice conversion. / Talsyntes är ett område som allt mer influerar vår vardag, exempelvis genom smarta assistenter, avancerade översättningssystem och liknande användningsområden. I det här examensarbetet utforskas fenomenet röstkonvertering, som innebär att man får en talare att låta som någon annan, utan att det som sades förändras. Mer specifikt undersöks ett Cycle-Consistent Adversarial Network som fungerat väl för röstkonvertering inom ett enskilt språk för röstkonvertering mellan olika språk. Det neurala nätverket tränas för konvertering mellan röster från olika modersmålstalare från de nordiska länderna. I experimenten används ingen parallell eller transkriberad data, vilket tvingar modellen att endast använda sig av ljudsignalen. Målet med examensarbetet är att utvärdera om modellens prestanda försämras i en flerspråkig kontext, jämfört med en enkelspråkig sådan, samt mäta hur stor försämringen i sådant fall är. I studien mäts prestanda i termer av kvalitet och talarlikhet för det genererade talet och rösten som efterliknas. För att utvärdera detta genomförs lyssningstester, samt objektiva analyser av det genererade talet. Resultaten visar att röstkonvertering mellan en svensk och norsk talare är möjlig utan att modellens prestanda försämras, jämfört med konvertering mellan svenska talare. För konvertering mellan finska och svenska talare, samt danska och svenska talare försämrades däremot kvaliteten av det genererade talet. Trots denna försämring producerade modellen tydligt och sammanhängande tal i samtliga experiment. Det här är anmärkningsvärt eftersom modellen tränades på mindre än 15 minuter icke-parallel data för varje talare. Detta examensarbete öppnar upp för nya framtida studier, exempelvis skulle fler språk kunna inkluderas eller nyare varianter av typen Generative Adversarial Network utvärderas. Mer resurser skulle även kunna läggas på att optimera hyperparametrarna för att ytterligare optimera den undersökta modellen för flerspråkig röstkonvertering.

Eyewitness identification plays an important role during criminal proceedings and may lead to prosecution and conviction of a suspect. One of the methods of eyewitness identification is a police photo lineup when a collection of photographs is presented to the witness in order to identify the perpetrator of the crime. In the lineup, there is typically at most one photograph (typically exactly one) of the suspect and the remaining photographs are the so-called fillers, i.e. photographs of innocent people. Positive identification of the suspect by the witness may result in charge or conviction of the suspect. Assembly of the lineup is a challenging and tedious problem, because the wrong selection of the fillers may end up in a biased lineup, where the suspect will stand out from the fillers and would be easily identifiable even by a highly uncertain witness. The reason why it is tedious is due to the fact that this process is still done manually or only semi-automatically. This thesis tries to solve both issues by proposing a model that will be capable of generating synthetic data, together with an application that will allow users to obtain the fillers for a given suspect's photograph. 1

No description available.

Analysis of overhead imagery has a great potential to produce real-time data cost-effectively. This can be an important foundation for decision-making for businesses and politics. Every day a massive amount of new satellite imagery is produced. To fully take advantage of these data volumes a computationally efficient pipeline is required for the analysis. This thesis proposes a pipeline which outperforms the Segment Before you Detect network [6] and different types of fast region based convolutional neural networks [61] with a large margin in a fraction of the time. The model obtains a prediction error for counting cars of 1.67% on the Potsdam dataset and increases the vehiclewise F1 score on the VEDAI dataset from 0.305 reported by [61] to 0.542. This thesis also shows that it is possible to outperform the Segment Before you Detect network in less than 1% of the time on car counting and vehicle detection while also using less than half of the resolution. This makes the proposed model a viable solution for large-scale satellite imagery analysis. / Analys av flyg- och satellitbilder har stor potential att kostnadseffektivt producera data i realtid för beslutsfattande för företag och politik. Varje dag produceras massiva mängder nya satellitbilder. För att fullt kunna utnyttja dessa datamängder krävs ett beräkningseffektivt nätverk för analysen. Denna avhandling föreslår ett nätverk som överträffar Segment Before you Detect-nätverket [6] och olika typer av snabbt regionsbaserade faltningsnätverk [61]  med en stor marginal på en bråkdel av tiden. Den föreslagna modellen erhåller ett prediktionsfel för att räkna bilar på 1,67% på Potsdam-datasetet och ökar F1- poängen for fordons detektion på VEDAI-datasetet från 0.305 rapporterat av [61]  till 0.542. Denna avhandling visar också att det är möjligt att överträffa Segment Before you Detect-nätverket på mindre än 1% av tiden på bilräkning och fordonsdetektering samtidigt som den föreslagna modellen använder mindre än hälften av upplösningen. Detta gör den föreslagna modellen till en attraktiv lösning för storskalig satellitbildanalys.

Yes / The early diagnosis and personalised treatment of diseases are facilitated by machine learning. The quality of data has an impact on diagnosis because medical data are usually sparse, imbalanced, and contain irrelevant attributes, resulting in suboptimal diagnosis. To address the impacts of data challenges, improve resource allocation, and achieve better health outcomes, a novel visual learning approach is proposed. This study contributes to the visual learning approach by determining whether less or more synthetic data are required to improve the quality of a dataset, such as the number of observations and features, according to the intended personalised treatment and early diagnosis. In addition, numerous visualisation experiments are conducted, including using statistical characteristics, cumulative sums, histograms, correlation matrix, root mean square error, and principal component analysis in order to visualise both original and synthetic data to address the data challenges. Real medical datasets for cancer, heart disease, diabetes, cryotherapy and immunotherapy are selected as case studies. As a benchmark and point of classification comparison in terms of such as accuracy, sensitivity, and specificity, several models are implemented such as k-Nearest Neighbours and Random Forest. To simulate algorithm implementation and data, Generative Adversarial Network is used to create and manipulate synthetic data, whilst, Random Forest is implemented to classify the data. An amendable and adaptable system is constructed by combining Generative Adversarial Network and Random Forest models. The system model presents working steps, overview and flowchart. Experiments reveal that the majority of data-enhancement scenarios allow for the application of visual learning in the first stage of data analysis as a novel approach. To achieve meaningful adaptable synergy between appropriate quality data and optimal classification performance while maintaining statistical characteristics, visual learning provides researchers and practitioners with practical human-in-the-loop machine learning visualisation tools. Prior to implementing algorithms, the visual learning approach can be used to actualise early, and personalised diagnosis. For the immunotherapy data, the Random Forest performed best with precision, recall, f-measure, accuracy, sensitivity, and specificity of 81%, 82%, 81%, 88%, 95%, and 60%, as opposed to 91%, 96%, 93%, 93%, 96%, and 73% for synthetic data, respectively. Future studies might examine the optimal strategies to balance the quantity and quality of medical data.