Contributions published at Department of Informatics (Burkhard Stiller)

With the increasing relevance of the risk concept in the field of information security, numerous new risk assessment methods have emerged. The selection of a suitable risk assessment method can prove itself to be a first obstacle for organizations that do not have the financial resources to employ consulting firms that assist with the risk assessment process. To address this challenge, comparison methods and tools have been developed in academia and the private sector. This thesis provides a comprehensive review of these methods and tools. An improved comparison method is designed based on this review and an in-depth analysis of several common information risk assessment methods. This method is then used to evaluate nine information security risk assessment methods. A navigable prototype for an information security risk assessment knowledge base has been designed and implemented, with the aim of facilitating comparison between methods and helping users select the most suitable assessment method. The improved comparison method is shown to be superior to predecessors by demonstrating that crucial criteria are adopted and novel criteria improve the selection process. Finally, a use case illustrates the efficiency of the prototype.

Point clouds are used in various algorithms in related fields, such as Visual Computing. Some of these algorithms require high-quality normals as input. Computing these surface normals is often the job of a different algorithm. Researchers have suggested vastly different algorithms over the years. Recent work focuses on using deep learning to estimate surface normals. It is important that these normal estimation algorithms perform well, as the other algorithms rely on their quality. This means the normal estimation algorithms need to be robust against various defects, including noise, outliers and differences in sampling density. Researches have proposed different measures to evaluate these normal estimation algorithms as well as how to produce synthetic test data to test against. Synthetic test data is especially useful as it provides ground truth normals to test against and can be constructed to contain a single or multiple defects with various strengths. Because of this, synthetic models are a valuable addition to real-world test data. In this paper, we evaluate four different normal estimation algorithms. This includes three regression based and one deep learning based method. We develop a tool to execute and compare the selected normal estimation algorithms. Our goal is to use good measures and cover multiple defects with our test data. In order to do this, we test the methods against various levels of noise and outliers. We use five different synthetic point cloud models and a real-world point cloud to test against. This allows us to identify the strengths and weaknesses of the tested methods.

Although influencer marketing has been experiencing tremendous growth, and many firms have been trying to find the right influencers for their marketing campaigns, there has been little research on identifying the timing of the high-impact work or analyzing the hot streak periods of influencers on social media. My thesis is among the first works in this area on the TikTok platform where I applied some methods used in previous literature. All the analyses were made using the video-level popularity metrics of TikTok: “number of likes”, “number of shares”, “number of plays”, and “number of comments”. In addition, unlike other studies, I created the visualization of the hot streak durations of TikTok authors to analyze information such as the start, and end time of hot streaks, or the length of hot streaks for each author. All plots for the analyses were built with bokeh- a Python library for interactive data visualizations. In conclusion, with the dataset in my thesis study, first I found that the timing of the most popular video is random in TikTok users’ lifecycles. Second, I discovered that the timing of the biggest hit and the second hit are close to each other, so TikTok authors may experience average success close to their most popular videos. Third, I detected a pattern indicating that TikTok authors use more diverse hashtags during hot streak periods than before them. Lastly, I observed that the relative hot streak length -success duration- of an author is usually between 12% and 50%. As a result, marketers can use these observations in my thesis to identify successful influencers on the TikTok platform. Finally, the formulas and visualization methods used in my thesis can be applied to a larger TikTok dataset or other datasets from other social media platforms such as Instagram, or YouTube.

Trust is a crucial component for successful transactions regardless of whether they are executed in physical or virtual spaces. Blockchain technology is often discussed in the context of trust and referred to as a trust-free, trustless, or trustworthy technology. However, the question of how the trustworthiness of blockchain platforms should be demonstrated and proven to end users still remains open. While there may be some genuine trust in the blockchain technology itself, on an application level trust in an IT artifact needs to be established. In this article, we examine how trust-supporting design elements may be implemented to foster an end user's trust in a blockchain platform. We follow the design science paradigm and suggest a practically useful set of design elements that can help designers of blockchain platforms to build more trustworthy systems.

Due to increased use cases in real-world applications, merging relational database management systems with linear algebra operations has been an ongoing topic. It allows the analysis of large amounts of data stored in database systems. Multiple approaches have been integrated, but the linear algebra operation, matrix transpose, remains particularly difficult to implement. This thesis attempts to allow the transposition of relations in database management systems and avoid previous issues encountered with matrix transposition. The solution is based on decoupling the logical and physical levels, in a database management system. Decoupling the two levels adds flexibility to the system and can be used to store relations differently than they are on their logical level. It was possible to directly implement the idea in the database management system MonetDB and evaluate it against a basic version of transpose. The evaluation shows some improvements in performance and the solution allows the transposition of relations with a large number of tuples, which has been a main issue for matrix transpose in database management systems.

The enormous growth of stored data continues to challenge our ability to efficiently process and analyze data. Many numerical computations related to state transition based on previous state, e.g., compound interest problems or newton’s method, require a combination of operations from the relational algebra (project, join, select etc.) and iterations. Moreover, many analytical computations, e.g., Markov chain algorithms or various types of regressions, require a combination of operations from the relational algebra, operations from the linear algebra, and iterations. Plenty of attempts have been made to solve combinations of those elements, but still improvement needed on time and space consumption. For example, some solutions focus on fetching data from the database, performing linear algebra operations and iterations, and then putting it back into the database, which lacks the support and optimization of relational algebra. Some solutions focus on linear algebra within the database, but require additional data structures and complicated preset functions, and the correspondence of non-numeric and numerical values is not defined. Moreover, the iteration of some solutions is very space or time consuming due to the operations involving table union and table updates. So, there is still not a very suitable solution that can perfectly combine the three aspects. Our iterations have ability to cover all three aspects at the same time, require no additional knowledge about Python, R, Hadoop or Spark, more tightly integrate with the classic SQL definition, and handle various types of iterations in more flexible and explicit ways. We describe the definition of our iterations in MonetDB, explain major design (motivated by various complicated iterations), and discuss key in-place features and future research directions. Finally, we provide applications and experiment results that show the potential of our solutions.

Newly released robot programming tools have made it feasible for end‐users to program industrial robots by combining block‐based languages and lead‐through programming. To use these systems effectively, end‐users, who usually have limited or no programming experience, require training. To train users, tutoring systems are often used for block‐based programming—some even for lead‐through programming—but no tutorial system combines these two types of programming. We present CoBlox Interactive Tutorials (CITs), a novel tutoring approach that teaches how to use both the hardware and software components that comprise a typical end‐user robot programming environment. As users switch between the two programming styles, CITs provide them with extensive scaffolding, give users immediate feedback on missteps, and provide guidance on next steps. To evaluate CITs, we conducted a study with 79 industrial end‐users using a programming environment released by ABB Robotics that compares our approach to training with training videos, the most commonly used training in industry. This study, one of the largest to date on training professional end‐users, found that CIT‐trained users authored more correct programs in less time than video‐trained users. This shows that a tight integration of hardware and software concepts is crucial to training end‐users to program industrial robots.

With the prosperity and development of the digital economy, many fraudsters have emerged on e-commerce platforms to fabricate fraudulent reviews to mislead consumers’ shopping decisions for profit. Moreover, in order to evade fraud detection, fraudsters continue to evolve and present the phenomenon of adversarial camouflage and collaborative attack. In this paper, we propose a novel temporal burstiness and collaborative camouflage aware method (TBCCA) for fraudster detection. Specifically, we capture the hidden temporal burstiness features behind camouflage strategy based on the time series prediction model, and identify highly suspicious target products by assigning suspicious scores as node priors. Meanwhile, a propagation graph integrating review collusion is constructed, and an iterative fraud confidence propagation algorithm is designed for inferring the label of nodes in the graph based on Loop Belief Propagation (LBP). Comprehensive experiments are conducted to compare TBCCA with state-of-the-art fraudster detection approaches, and experimental results show that TBCCA can effectively identify fraudsters in real review networks with achieving 6%–10% performance improvement than other baselines.

The connectivity and resource-constrained nature of single-board devices open the door to cybersecurity concerns affecting Internet of Things (IoT) scenarios. One of the most important issues is the presence of unauthorized IoT devices that want to impersonate legitimate ones by using identical hardware and software specifications. This situation can provoke sensitive information leakages, data poisoning, or privilege escalation in IoT scenarios. Combining behavioral fingerprinting and Machine/Deep Learning (ML/DL) techniques is a promising approach to identify these malicious spoofing devices by detecting minor performance differences generated by imperfections in manufacturing. However, existing solutions are not suitable for single-board devices since they do not consider their hardware and software limitations, underestimate critical aspects such as fingerprint stability or context changes, and do not explore the potential of ML/DL techniques. To improve it, this work first identifies the essential properties for single-board device identification: uniqueness, stability, diversity, scalability, efficiency, robustness, and security. Then, a novel methodology relies on behavioral fingerprinting to identify identical single-board devices and meet the previous properties. The methodology leverages the different built-in components of the system and ML/DL techniques, comparing the device internal behavior with each other to detect variations that occurred in manufacturing processes. The methodology validation has been performed in a real environment composed of 15 identical Raspberry Pi 4 Model B and 10 Raspberry Pi 3 Model B＋ devices, obtaining a 91.9% average TPR with an XGBoost model and achieving the identification for all devices by setting a 50% threshold in the evaluation process. Finally, a discussion compares the proposed solution with related work, highlighting the fingerprint properties not met, and provides important lessons learned and limitations.

We analyze short-term media trust changes during the COVID-19 pandemic, their ideological drivers and consequences based on panel data in German-speaking Switzerland. We thereby differentiate trust in political information from different types of traditional and non-traditional media. COVID-19 serves as a natural experiment, in which citizens’ media trust at the outbreak of the crisis is compared with the same variables after the severe lockdown measures were lifted. Our data reveal that (1) media trust is consequential as it is associated with people’s willingness to follow Covid-19 regulations; (2) media trust changes during the pandemic, with trust levels for most media decreasing, with the exception of public service broadcasting; (3) trust losses are hardly connected to ideological divides in Switzerland. Our findings highlight that public service broadcasting plays an exceptional role in the fight against a pandemic and that contrary to the US, no partisan trust divide occurs.

Traffic accidents are the leading cause of death among young people, a problem that today costs an enormous number of victims. Several technologies have been proposed to prevent accidents, being brain–computer interfaces (BCIs) one of the most promising. In this context, BCIs have been used to detect emotional states, concentration issues, or stressful situations, which could play a fundamental role in the road since they are directly related to the drivers’ decisions. However, there is no extensive literature applying BCIs to detect subjects’ emotions in driving scenarios. In such a context, there are some challenges to be solved, such as (i) the impact of performing a driving task on the emotion detection and (ii) which emotions are more detectable in driving scenarios. To improve these challenges, this work proposes a framework focused on detecting emotions using electroencephalography with machine learning and deep learning algorithms. In addition, a use case has been designed where two scenarios are presented. The first scenario consists in listening to sounds as the primary task to perform, while in the second scenario listening to sound becomes a secondary task, being the primary task using a driving simulator. In this way, it is intended to demonstrate whether BCIs are useful in this driving scenario. The results improve those existing in the literature, achieving 99% accuracy for the detection of two emotions (non-stimuli and angry), 93% for three emotions (non-stimuli, angry and neutral) and 75% for four emotions (non-stimuli, angry, neutral and joy).

Automatic medical text simplification can assist providers with patient-friendly communication and make medical texts more accessible, thereby improving health literacy. But curating a quality corpus for this task requires the supervision of medical experts. In this work, we present MedEASi (Medical dataset for Elaborative and Abstractive Simplification), a uniquely crowdsourced and finely annotated dataset for supervised simplification of short medical texts. Its expert-layman-AI collaborative annotations facilitate controllability over text simplification by marking four kinds of textual transformations: elaboration, replacement, deletion, and insertion. To learn medical text simplification, we fine-tune T5-large with four different styles of inputoutput combinations, leading to two control-free and two controllable versions of the model. We add two types of controllability into text simplification, by using a multi-angle training approach: position-aware, which uses in-place annotated inputs and outputs, and position-agnostic, where the model only knows the contents to be edited, but not their positions. Our results show that our fine-grained annotations improve learning compared to the unannotated baseline. Furthermore, position-aware control generates better simplification than the position-agnostic one. The data and code are available at https://github.com/Chandrayee/CTRL-SIMP.

Trusting Artificial Intelligence (AI) is controversial since models and predictions might not be fair, understandable by humans, robust against adversaries, or trained appropriately. Existing toolkits help data scientists to create fair, explainable, robust, and transparent Machine and Deep Learning (ML/DL) models. However, tools to quantify AI trustworthiness according to pillars and metrics relevant for heterogeneous scenarios are still missing. This work proposes a novel algorithm that quantifies the trustworthiness level of supervised ML/DL models according to their fairness, explainability, robustness, and accountability. The algorithm is deployed on a Web application to allow the general public to calculate the trustworthiness of their models. Finally, a validation scenario with models classifying cyberattacks demonstrates the applicability of the Web application and algorithm.

Information communication technology (ICT)’s environmental impact must be considered in digital innovation and associated policy to mitigate ICT’s climate change contribution. A pro- posed solution to reduce ICT emissions is by improving efficiency, yet this fails to consider rebound effects where efficiency improvements offset emissions savings or increase emissions. In this perspective, we reveal insights from a transdisciplinary workshop that identified challenges for why rebound effects are difficult to include in innovation and policy. From this, we call researchers to (1) find new ways of presenting rebound effects to digital innovators and policymakers; (2) gather cross-disciplinary evidence of ICT rebound effects; and (3) transparently analyze ICT’s environmental, societal, and economic impacts together. We also call for a systems thinking approach to addressing ICT’s environmental impacts, whereby a solution to rebound ef- fects becomes visible: efficiencies under emission constraints.

Ranking the spreading influence of nodes is of great importance in practice and research. The key to ranking a node’s spreading ability is to evaluate the fraction of susceptible nodes being infected by the target node during the outbreak, i.e. the outbreak size. In this paper, we present a dynamic Markov process (DMP) method by integrating the Markov chain and the spreading process to evaluate the outbreak size of the initial spreader. Following the idea of the Markov process, this method solves the problem of nonlinear coupling by adjusting the state transition matrix and evaluating the probability of the susceptible node being infected by its infected neighbors. We have employed the susceptible-infected-recovered and susceptible-infected-susceptible models to test this method on real-world static and temporal networks. Our results indicate that the DMP method could evaluate the nodes’ outbreak sizes more accurately than previous methods for both single and multi-spreaders. Besides, it can also be employed to rank the influence of nodes accurately during the spreading process.

We study the combinatorial assignment domain, which includes combinatorial auctions and course allocation. The main challenge in this domain is that the bundle space grows exponentially in the number of items. To address this, several papers have recently proposed machine learning-based preference elicitation algorithms that aim to elicit only the most important information from agents. However, the main shortcoming of this prior work is that it does not model a mechanism's uncertainty over values for not yet elicited bundles. In this paper, we address this shortcoming by presenting a Bayesian optimization-based combinatorial assignment (BOCA) mechanism. Our key technical contribution is to integrate a method for capturing model uncertainty into an iterative combinatorial auction mechanism. Concretely, we design a new method for estimating an upper uncertainty bound that can be used to define an acquisition function to determine the next query to the agents. This enables the mechanism to properly explore (and not just exploit) the bundle space during its preference elicitation phase. We run computational experiments in several spectrum auction domains to evaluate BOCA's performance. Our results show that BOCA achieves higher allocative efficiency than state-of-the-art approaches.

Considering the condition of our planet, anticipating natural disasters has long been a hot topic. This work is becoming more doable thanks to the expansion of earth observation data sources, such as satellite imagery. In this work, our main interest is droughts and their impacts. Recent hot and dry summers in Europe have had a significant impact on forest functioning and structure. In 2018 and 2019, Central Europe experienced two extremely dry and hot summers. These extremes resulted in widespread canopy defoliation and tree mortality. The objective of this study is to create a predictive model for forecasting future satellite imagery that contains information about the greenness of vegetation as measured by the Normalized Difference Vegetation Index (NDVI). We predict NDVI utilising data from the previous months as input to determine where and when drought impacts are triggered. We use a combination of temporal bands from Sentinel 2 and ERA-5 data sources, as well as static data sources such as the NASA SRTM Digital Elevation Model and the Copernicus Landcover Classification Map, as predictors. We will now focus on the forests of Switzerland as a region of interest in order to leverage high-quality model input layers and applications to meet typical stakeholder needs. Widely used vegetation indices and mechanistic land surface models are not effectively informed by the full information contained in Earth observation data and the observed spatial heterogeneity of land surface greenness responses at hillslope-scale resolution. Effective learning from the simultaneous evolution of climate and remotely sensed land surface properties is challenging. Modern deep learning and machine learning techniques, however, have the capacity to generate accurate predictions while also explaining the relationship between climate and its recent history, its position in the landscape, and its influences on vegetation. The task is to predict the future NDVI over forest areas to infer droughts, given past and future weather and surface reflectance. Giving future weather predictions as an input to the model, we are going for a 'guided prediction' approach where the aim is to exploit weather information from forecasting models in order to increase the predictive power of the model. Models are fully data-driven, without feature engineering, and trained on spatio-temporal data cubes, which can be seen as stacked satellite imagery for a specific geo-location and a timestamp of past Sentinel 2 surface reflectance, past (observed) and future (forecasted) climate reanalysis, time-invariant information from a digital elevation model, and a land cover map. In the temporal domain, models are trained on the period between 2018-2019, validated between 05/2021 and 09/2021, and tested between 05/2020 and 09/2020.  In this research, we propose a methodology for how to successfully integrate future data from different modalities to go for a "guided-prediction" approach to enhance the predictive power of the models. We also propose a novel, complete guideline for how to effectively create earth observation data cubes. We conducted experiments regarding the model's performance under sparse conditions (clouds). We observed that the proposed model out-performed the baseline. However instead of learning the true signal, model "memorised" of the imputation values used to replace cloudy pixel values. We believe that the reasons for this are the small amount of data to learn from, which effects the generalizability skill of the model, and our chosen cloud removal strategy.

Semantic segmentation is an important task in computer vision. It performs pixel-level labeling with a set of object categories (e.g., human, car, tree, sky) for all image pixels; thus, it is generally a more demanding undertaking than whole-image classification, which predicts a single label for the entire image. Since Machine Learning is proposed, numerous supervised models have achieved very good performance in semantic segmentation tasks with reasonable computation costs. However, the performance of the supervised model is limited by the quality and amount of the labeled datasets, which are scarce and expensive to obtain. This work adapts a popular semi-supervised learning method, namely consistency learning, to the retinal vessel segmentation task. The main idea of this method is to minimize the differences between two predictions generated from two variants, which are produced by applying data augmentations to the same input, meanwhile, to maximize the agreement between the prediction and the ground truth. Because the distribution of pixels belonging to the vessels is sparse, limited data augmentations can be applied to the samples to produce the variants in this task. We figure out the basic data augmentations providing the best performance and test the model on four publicly available datasets. Our results suggest that our model can significantly improve the prediction performance on the labeled/unlabeled dataset pairs which have poor generalization ability in the supervised learning methods. For an unseen dataset, it is important to choose the labeled dataset used in training carefully. When the model is trained with a properly chosen labeled dataset, increasing the number of unlabeled datasets can improve its performance.

A fundamental task in many modern task-oriented dialogue systems is intent classification, in which the user's text input is mapped to a predefined intent. However, task-oriented dialog systems support a limited number of intents, and a key challenge they face is to reject unknown intents. Open-set recognition aims to solve this problem of classifying known classes correctly and rejecting the unknown. One way to train models to reject unknowns is to include representatives of unknown classes during training, called negative samples. In this thesis, we propose several approaches for synthetic negative sample generation to improve model performance on open-set recognition. We first extend the Manifold Mixup approach with different sample selection strategies and apply it to different layers of the network. We also propose using adversarial text attack samples as another source of negative samples. In addition, we apply Entropic Open-Set (EOS) loss function that was shown to improve open-set recognition performance on images. Our experiments compare these approaches with baseline approaches using Open-set Classification Rate (OSCR) curve that was proposed specifically for the open-set recognition task. Our results show that negative samples from adversarial attacks on text could be effective for open-set recognition in certain scenarios. On the other hand, Manifold Mixup-based approaches, including a state-of-the-art approach, are on par with the baselines considering the trade-off between correctly classifying known samples and rejecting unknown samples.