Contributions published at Communication Systems Group (Burkhard Stiller)

The right to vote for laws and political positions is a central part of modern democracy. While voting through post or at the voting site are still the most prevalent forms of submitting one's ballot, there exist recent efforts in taking this process online. Remote Electronic voting (REV) systems allow the voting population to submit their votes from the comfort of their own home, through internet-capable devices such as personal computers or smartphones. Casting a vote is a process of high severity, and for a lot of persons, bears emotional attachment. Recent literature and user studies on REV systems have uncovered concerns about security and privacy within the voters' mental models of casting their votes online. In most cases these concerns are covered by the software and thus don't actually form a threat to a vote's integrity. However, the system's user interface (UI) may not elicit the confidence and trust required for using the application properly while providing a positive experience. A blockchain (BC) offers several desirable properties that are striven for in a REV system such as immutability, decentralization, and transparency. However, BCs are also technical concepts for which a typical voter does not have a mental model. Thus for any REV system using BC technology, it is important to take special care with its UI design. Provotum is a BC-based REV system developed as a research project at the University of Zurich. Its latest version (3.0) received a technical upgrade and security audit, improving security, scalability, and introducing receipt-freeness. This version needs to be orchestrated in a headless fashion, meaning that there is no graphical user interface (GUI) to operate it. This thesis analyses literature in the fields of UI and user experience (UX) design both within the domain of REV systems as well as in the general realm of modern software systems. Furthermore, the usability, and lack thereof, of Provotum 3.0 is analyzed. Based on these findings, UI designs for the vote administration software and the voting software are proposed. The designs are prototypically implemented using state-of-the-art technology. Both applications are evaluated based on well-established heuristic rules as well as based on the coverage of usage scenarios.

The heterogeneous nature of raw milk production, transportation, processing, transformation, and storage have been triggering concerns on the authenticity of dairy products. For the first time known, this work studied these concerns via interviews with consumers and producers of dairy products. Therefore, this paper introduces “NUTRIA” as a decentralized dairy product Supply Chain Tracing (SCT) system, designed and implemented based on real-world observations of the Swiss dairy supply chain and conducted in collaboration with dairy producers. Based on these studies and to overcome deficits of traditional and centralized SCT approaches, NUTRIA enables an automated SCT via a Blockchain-based decentralized application. NUTRIA materializes a trusted and transparent SCT, which empowers the dairy value chain. Furthermore, results of the real-world evaluations of deployed approaches within NUTRIA are covered, including social aspects and risks.

null

null

null

null

null

Continuous authentication systems have been proposed as a promising solution to authenticate users in smartphones in a non-intrusive way. However, current systems have important weaknesses related to the amount of data or time needed to build precise user profiles, together with high rates of false alerts. Voice is a powerful dimension for identifying subjects but its suitability and importance have not been deeply analyzed regarding its inclusion in continuous authentication systems. This work presents the S3 platform, an artificial intelligence-enabled continuous authentication system that combines data from sensors, applications statistics and voice to authenticate users in smartphones. Experiments have tested the relevance of each kind of data, explored different strategies to combine them, and determined how many days of training are needed to obtain good enough profiles. Results showed that voice is much more relevant than sensors and applications statistics when building a precise authenticating system, and the combination of individual models was the best strategy. Finally, the S3 platform reached a good performance with only five days of use available for training the users’ profiles. As an additional contribution, a dataset with 21 volunteers interacting freely with their smartphones for more than sixty days has been created and made available to the community.

In recent years the blockchain topic, the underlying technology of cryptocurrencies, has gained increasing importance and attention. Since the invention of Bitcoin in 2009, the number of cryptocurrencies and various blockchain platforms has drastically increased. With such a myriad of implementations and the resulting lack of transparency, it become complex to select a suitable blockchain for a specific use case. Recently, a policy-based management approach has been proposed to automate the selection process, which recommends blockchain implementations based on transaction information and pre-defined policies. This selection process is governed by a simple algorithm, which applies straightforward filtering. The goal of this thesis is the research of novel approaches, such as machine learning, and apply them to the blockchain selection process with the aim of integrating them into the existing solution. Therefore, various machine learning algorithms were trained, deployed and evaluated on their applicability in the blockchain selection process. The developed prototype extends the existing solution and provides users the option to choose between the conventional and the machine learning-based selection algorithm, with changing policy parameters depending on the selection. The results of the evaluation show that such a combination is in fact feasible and can deliver accurate results. But it also highlights pitfalls and necessary considerations.

The authentication field is evolving towards mechanisms able to keep users continuously authenticated without the necessity of remembering or possessing authentication credentials. While relevant limitations of continuous authentication systems -high false positives rates (FPR) and difficulty to detect behaviour changes- have been demonstrated in realistic single-device scenarios, the Internet of Things and next generation of mobile networks (5G) are enabling novel multi-device scenarios, such as Smart Offices, that can help to reduce or address the previous challenges. The paper at hand presents an AI-based, privacy-preserving and multi-device continuous authentication architecture called AuthCODE. AuthCODE seeks to improve single-device solutions limitations by considering additional behavioural data coming from heterogeneous devices. AuthCODE proposes a novel set of features that combine the interactions of users with different devices. The features relevance has been demonstrated in a realistic Smart Office scenario with several users that interact with their mobile devices and personal computers. In this context, a set of single- and multi-device datasets have been generated and published to compare the performance of our multi-device solution against single-device approaches. A pool of experiments with machine and deep learning classifiers measured the impact of time in authentication accuracy and improved the results of single-device approaches by considering multi-device behaviour profiles. Specifically, the multi-device approach using XGBoost with 1-minute window of aggregated features, achieved a 69.33%, 59,65% and 89,35% improvement in the FPR when compared to the single-device approach for computer, mobile applications and mobile sensors respectively. Finally, temporal information classified by a Long-Short Term Memory Network, allowed the identification of additional complex behaviour patterns.

Fair, secure and trustworthy voting processes and elections are a cornerstone of any functioning democracy. Caused by the ongoing digitalization and digitization, new ways of voting are emerging. A highly discussed and promising topic is remote electrical voting, which allows users to cast their votes independent of the location and additionally (partly) independent of the used device. The combination of this would bring more flexibility to voters since they must not be present at a poll station at a particular time. For countries already allowing voting by mail, remote electronic voting is an evolution of this known practice. As with every new technology, also with remote electrical voting, many obstacles must be conquered before it will be secure and reliable enough to be rolled out to the public. Especially in the scenario of electronic voting, cryptographic operations and encryption must be applied to guarantee features like vote secrecy, the resistance against coercion, or prevention against vote selling. Nevertheless, a voter must be condident that the indeed selected voting option is encrypted, transferred to the ballot box, and later counted in the tally. Proving this to the voter is not trivially since most electrical and cryptographic operations performed on data are not comprehensible and not verifiable without proper auxiliary tools. This work focuses on the verification of the encryption of a selected voting option. The verification allows voters to verify if a ballot contains the chosen voting option or if the voting device tampers the selection before encrypting it. This will enable voters to verify whether their voting setup encrypts the selection trustworthy or if the voting device is cheating and altering the selection. Hence, voting with an unknown, unfamiliar, or not trusted device is possible. The thesis shows how to successfully implement the cast-as-intended property with the help of the challenge-or-cast mechanism into an existing remote electronic voting system. The challenge-or-cast mechanism allows voters to either challenge the encryption of a ballot or cast it. For the verification, a second device is needed such that the encryption can be repeated in an air-gapped environment. Furthermore, the challenge-or cast approach is compared to other mechanisms having the same goal. Towards the end, the selected method is analyzed and discussed, revealing strengths, weaknesses, chances, and concerns.

Since many years, traditional voting schemes have been challenged by e-voting systems. However, e-voting systems are only actively used in some areas of the world, since proposed systems are often error-prone or do not meet the necessary security requirements. The Swiss remote postal voting is very complex (for the voter) and is heavily based on the voter's trust in third parties. This work tries to make the current voting system more verifiable by adapting small components of the physical voting material. Specifically, this means that additional QR codes are printed on the voting card. These QR codes contain an encrypted entry of the electoral register concerning the respective voter. The corresponding encrypted entries are made accessible through IPFS on an authorized website and can be compared with the received QR code. In addition, every PB contains an RFID tag. With the help of this tag, the electoral officer can reproduce the right path of the paper ballot without revealing the identity of the voter. Overall, nothing is changed in the weighting between privacy and verifiability. The purpose of the suggested scheme is to be more verifiable while staying receipt-free. However, since the use of RFID technology causes additional costs, it remains an open question how applicable the proposed scheme is.

With the introduction of the concept of blockchain (BC) and the subsequent launch of Bitcoin in 2009, a trend of rapidly increasing numbers of alternative BC implementations and platforms has begun. Due to their inherent differences, different BCs are generally incapable of exchanging neither their native currencies nor the data stored on them. Thus, numerous interoperability solutions have been envisioned, with Bifröst, a notary-scheme based interoperability API, being one of them. Bifröst's prototype implementation, though working as intended, had a number of key-areas where improvements were possible, thus, in this thesis, new features have been added to strengthen the core requirements imposed on Bifröst. Optional data encryption has been incorporated to give users more control and to increase the security of stored confidential information, at the cost of increased data sizes when encrypted. The capability to have oversized data split, stored with multiple transactions and tracked properly for reassembly on retrieval has been included, increasing the flexibility of the data that can be stored. Furthermore, ease of use has been improved with the addition of generic error handling, whilst at the same time, in conjunction with redundancy, increasing the robustness of Bifröst. Research on standardized interoperability formats has been conducted and has served as inspiration for a new JSON scheme for interacting with BC interoperability APIs such as Bifröst. Finally, means to securely manage private keys have been investigated, and although ultimately no immediately actionable solution was found, the options for future developments of Bifröst in that area have been clarified.

Due to advancing digitalization, the continuous growth of e-commerce and the increasing acceptance of mobile and contactless payment methods, an increasing number of purchases are being made with credit and debit cards [3]. However, this growth has its downsides. Where there is a lot of money at stake, fraudsters are not far away. The study conducted by [1] mentions that the global annual loss of 28 billion due to fraud will rise to over 35 billion within 5 years. To counter this trend, credit card issuers are investing heavily in credit card fraud detection and prevention systems. In the industry, mainly rule-based systems [2] are currently in operation. There is a strong trend towards machine learning models as they offer a number of advantages over static rules. However, a machine learning model requires extensive and sophisticated data preparation. This is further complicated by the fact that each card usage has to be evaluated within a few milliseconds. Traditional systems reach their limits in this preparation task. Streaming technology could provide a remedy. Accordingly, this thesis deals with the design and implementation of algorithms and heuristics to optimize a streaming system for credit card fraud detection and prevention. Different optimization strategies are introduced to increase the performance of a given system. Thereby, a subdivision into operational state, Input/Output (I/O) and algorithmic optimizations takes place. To the best of the authorís knowledge, the cascading window aggregation and the continuous sliding window algorithm are introduced as two new optimization approaches. With the implemented adjustments, the throughput of the streaming job could be improved from a few 100 to 40í000 events per second. Furthermore, the job is no longer clogged and the status does not longer get bloated. By reducing the number of operators required, the latency could also be significantly reduced.

With the emergence of several thousand Blockchains in recent years, the selection of use-case appropriate Blockchains has become a formidable task. Withal, hitherto no Blockchain standardization body is prevalent, leading to a research gap concerning the characterization and classification of Blockchains. Therefore, this thesis' approach to bridge this gap is by taking inspiration from software engineering principles, namely Functional and Non-Functional Requirements and their characterization using Softgoal Interdependency Graphs. Through decompositions and quantification of relationships, these graphs allow the understanding of how certain Blockchain attributes and aspects are achieved and obtained, and they facilitate the comparison of various Blockchain implementations. Consequently, a Blockchain specific Softgoal Interdependency Graph was designed and evaluated on two relevant use cases. Due to the inherent structure of the graph, a Blockchain characterization relating to four different attributes was achieved, and a Machine Learning evaluation on the use cases resulted in a classification of three different Blockchain clusters. Therefore, this approach can be deemed as successful, and future work can utilize this process and the resulting values to incorporate them into the Blockchain Selection task.

Web platforms are gaining relevance in eHealth, where they ease the interaction between patients and clinician. However, some clinical fields, such as the cardiovascular one, still need more effort because cardiovascular diseases are the principal cause of death and medical resources expenditure worldwide. The lack of daily control is the main reason hypertension is a current health problem, and medical web services could improve this situation. To face this challenge, this work proposes a novel intelligent web-based ecosystem, called SENIOR, capable of predicting adverse blood pressure events. The innovation of the SENIOR ecosystem relies on a wearable device measuring patient’s biomarkers such as blood pressure, a mobile application acquiring patient’s information, and a web platform consulting environmental services, processing data, and predicting blood pressure. The second contribution of this work is to consider novel environmental features based on the users’ location, such as climate and pollution data, to increase the knowledge about known variables affecting hypertension. Finally, our last contribution is a proof of concept with several machine learning algorithms predicting blood pressure values both in real-time and future temporal windows within one day has demonstrated the suitability of SENIOR.

In the current network-based computing world, where the number of interconnected devices grows exponentially, their diversity, malfunctions, and cybersecurity threats are increasing at the same rate. To guarantee the correct functioning and performance of novel environments such as Smart Cities, Industry 4.0, or crowdsensing, it is crucial to identify the capabilities of their devices (e.g., sensors, actuators) and detect potential misbehavior that may arise due to cyberattacks, system faults, or misconfigurations. With this goal in mind, a promising research field emerged focusing on creating and managing fingerprints that model the behavior of both the device actions and its components. The article at hand studies the recent growth of the device behavior fingerprinting field in terms of application scenarios, behavioral sources, and processing and evaluation techniques. First, it performs a comprehensive review of the device types, behavioral data, and processing and evaluation techniques used by the most recent and representative research works dealing with two major scenarios: device identification and device misbehavior detection. After that, each work is deeply analyzed and compared, emphasizing its characteristics, advantages, and limitations. This article also provides researchers with a review of the most relevant characteristics of existing datasets as most of the novel processing techniques are based on Machine Learning and Deep Learning. Finally, it studies the evolution of these two scenarios in recent years, providing lessons learned, current trends, and future research challenges to guide new solutions in the area.

The reliable measurement of objects in time and space is a prerequisite for many future technologies such as autonomous driving. There are various methods based on different measuring instruments. One of these measurement methods for distance measurement and object gathering is called light detection and ranging, or LiDAR for short. For example, Volvo, known for its high vehicle safety, is using LiDAR sensors in its next generation of vehicles. But not only movements of road users are interesting but also movements in a room, for example to determine the visitor frequency at an exhibition. This bachelor thesis is about the design of a system for tracking static and moving objects, in this case people, in a closed room with the help of a LiDAR sensor. For this purpose an interface between a LiDAR scanner and a database is developed. In the independent, subsequent data processing the data is checked for objects. The developed system called LaFlector can detect, classify and track several objects simultaneously. The detected objects are recorded and dynamically displayed to the user in a coordinate system. The evaluation shows that the system can reliably detect, classify and track objects, taking into account the limitations of a single LiDAR scanner.

In order to provide a solution that aids in analyzing and detecting ever-increasing Distributed Denial of Service (DDoS) attacks in post, DDoSGrid was created, which offers a research-oriented and extensible platform that provides visualizations for analyzed traffic logs and lets a user configure dashboards using these visualizations. However, this analysis and attack-type detection was until now performed manually, by a user looking at visualizations and metrics and making an informed decision on what attack vectors could have been used in an attack. This thesis expands on the base-idea of DDoSGrid and extends it by conceptualizing and implementing an extension that allows machine learning based attack-type classification of the uploaded data sets. In addition to the already existing feature extractors that DDoSGrid provides, a new extractor is created that creates time-window based features of the traffic log. These logs can be manually classified, and then added to a machine learning model, to create a true data set. This model is then used to automatically classify new data sets using different classification algorithms, in order to get an attack-type analysis of the traffic log. The solution was evaluated using well-established techniques and proved to be quite effective, both in terms of performance and accuracy, performing comparably or in some cases better than the existing literature.