000 Informatik, Informationswissenschaft, allgemeine Werke
Refine
Year of publication
Document Type
- Doctoral Thesis (24)
- Article (8)
- Conference Proceeding (1)
- Report (1)
Has Fulltext
- yes (34)
Is part of the Bibliography
- no (34)
Keywords
- Privatsphäre (2)
- Abfragesprache (1)
- Active Learning (1)
- Allianz Aktiengesellschaft (1)
- Alltagsdigitalisierung (1)
- Anfragerelaxation (1)
- Armut (1)
- Artificial Intelligence (1)
- Avionik (1)
- Bildverarbeitung (1)
With the increasing use of digital technologies in the automotive sector, the traditional automobile is undergoing a structural transformation, requiring new technologies and enabling innovative mobility concepts.
In particular, the ability to drive automatically or even fully autonomously, update control software, and remain connected to the environment allows attackers to infiltrate highly critical vehicle systems and take control without adequate protection.
Once not only individual vehicles but entire fleets are dominated by software, cyberattacks could disrupt a significant portion of the infrastructure and expose passengers to substantial risks.
This work follows a holistic approach to protecting highly automated software-defined vehicles from cyberattacks by designing and implementing security concepts in the main phases of a vehicle's lifecycle.
We use SAE level 4 prototype vehicles to evaluate our proposed techniques.
We start with a systematic security requirement analysis using the ISA-62443 standard series, demonstrating how threats can be identified in a collaborative, hierarchical process and how the resulting security risks impact the software and hardware architecture of a self-driving vehicle.
We show how this analysis process results in concrete requirements whose consideration reduces the overall security risk to a tolerable level.
Subsequently, we develop technical solutions for selected requirements. We begin by securing the CAN and FlexRay legacy protocols, which we foresee being used in specific areas of SDV in a transitional period despite technological changes.
To enable vehicle-wide security management, we address the management and distribution of cryptographic keys within such networks, mainly focusing on resource-constrained devices.
We propose using lightweight implicit certificates for deriving cryptographic group keys that can be used in CAN networks.
Additionally, we demonstrate how the slot-based frame structure of the FlexRay protocol allows for efficient "multi-slot" authentication, for which we calculate cryptographic keys using hash-based key chains.
SDV use Ethernet-based communication protocols and custom middleware stacks to transmit large amounts of data in real-time.
We develop a three-stage security process for the novel ASOA, which enables the development and central orchestration of system-agnostic functional software components on embedded systems and HPC platforms.
After the central specification of the security architecture at the data flow level, security tokens are automatically calculated and distributed for runtime protection of the service-oriented, DDS-based data transmission.
Our process ensures the strict separation of function and system knowledge, allowing for cost-effective and adaptable security architecture management.
The evaluation in four self-driving, software-defined vehicles demonstrates an average runtime overhead of approximately 5.71%.
As the initial risk analysis and actual cyberattacks have shown, protective measures against the compromise of control units must be taken alongside communication security.
To address this, we develop a method for verifying and validating the software integrity of control units.
A governmental third party confirms a measurement through a digital certificate, proving the examined vehicle's trustworthiness and suitability for participation in automated traffic.
In the final step of this work, we present an assessment scheme that allows software-defined vehicles to evaluate security incidents during operation in terms of their maximum expected damage and initiate appropriate countermeasures.
We follow the ISO/SAE 21434 standard and model attack paths using a graph representing dependencies among internal vehicle assets to account for the propagation effects of cyberattacks.
The assessment of a security incident considers not only the probability of individual attack paths but also the vehicle context.
Our practical evaluation demonstrates that we can detect, report, and assess security incidents below the human reaction time in the earlier mentioned prototype vehicles.
The collection of personal information by organizations has become increasingly essential for social interactions. Nevertheless, according to the GDPR (General Data Protection Regulation), the organizations have to protect collected data. Access Control (AC) mechanisms are traditionally used to secure information systems against unauthorized access to sensitive data. The increased availability of personal sensor data, thanks to IoT-oriented applications, motivates new services to offer insights about individuals. Consequently, data mining algorithms have been proposed to infer personal insights from collected sensor data. Although they can be used for genuine purposes, attackers can leverage those outcomes, combining them with other type of data, and further breaching individuals’ privacy. Thus, bypassing AC mechanisms thanks to such insights is a concrete problem.
We propose an inference detection system based on the analysis of queries issued on a sensor database. The knowledge obtained through these queries, and the inference channels corresponding to the use of data mining algorithms on sensor data to infer individual information, are described using Raw sensor data based Inference ChannEl Model (RICE-M). The detection is carried out by RICE-M based inference detection System (RICE-Sy). RICE-Sy considers at the time of the query, the knowledge that a user obtains via a new query and has obtained via his query history, and determines whether this is sufficient to allow that user to operate a channel. Thus, privacy protection systems can take advantage of the inferences detected by RICE-Sy, taking into account individuals’ information obtained by the attackers via a database of sensors, to further protect these individuals.
Vanadium redox-flow batteries (VRFBs) have played a significant role in hybrid energy storage systems (HESSs) over the last few decades owing to their unique characteristics and advantages. Hence, the accurate estimation of the VRFB model holds significant importance in large-scale storage applications, as they are indispensable for incorporating the distinctive features of energy storage systems and control algorithms within embedded energy architectures. In this work, we propose a novel approach that combines model-based and data-driven techniques to predict battery state variables, i.e., the state of charge (SoC), voltage, and current. Our proposal leverages enhanced deep reinforcement learning techniques, specifically deep q-learning (DQN), by combining q-learning with neural networks to optimize the VRFB-specific parameters, ensuring a robust fit between the real and simulated data. Our proposed method outperforms the existing approach in voltage prediction. Subsequently, we enhance the proposed approach by incorporating a second deep RL algorithm—dueling DQN—which is an improvement of DQN, resulting in a 10% improvement in the results, especially in terms of voltage prediction. The proposed approach results in an accurate VFRB model that can be generalized to several types of redox-flow batteries.
The worldwide adoption of Electric Vehicles (EVs) has embraced promising advancements toward a sustainable transportation system. However, the effective charging scheduling of EVs is not a trivial task due to the increase in the load demand in the Charging Stations (CSs) and the fluctuation of electricity prices. Moreover, other issues that raise concern among EV drivers are the long waiting time and the inability to charge the battery to the desired State of Charge (SOC). In order to alleviate the range of anxiety of users, we perform a Deep Reinforcement Learning (DRL) approach that provides the optimal charging time slots for EV based on the Photovoltaic power prices, the current EV SOC, the charging connector type, and the history of load demand profiles collected in different locations. Our implemented approach maximizes the EV profit while giving a margin of liberty to the EV drivers to select the preferred CS and the best charging time (i.e., morning, afternoon, evening, or night). The results analysis proves the effectiveness of the DRL model in minimizing the charging costs of the EV up to 60%, providing a full charging experience to the EV with a lower waiting time of less than or equal to 30 min.
ChatGPT and similar generative AI models have attracted hundreds of millions of users and have become part of the public discourse. Many believe that such models will disrupt society and lead to significant changes in the education system and information generation. So far, this belief is based on either colloquial evidence or benchmarks from the owners of the models—both lack scientific rigor. We systematically assess the quality of AI-generated content through a large-scale study comparing human-written versus ChatGPT-generated argumentative student essays. We use essays that were rated by a large number of human experts (teachers). We augment the analysis by considering a set of linguistic characteristics of the generated essays. Our results demonstrate that ChatGPT generates essays that are rated higher regarding quality than human-written essays. The writing style of the AI models exhibits linguistic characteristics that are different from those of the human-written essays. Since the technology is readily available, we believe that educators must act immediately. We must re-invent homework and develop teaching concepts that utilize these AI models in the same way as math utilizes the calculator: teach the general concepts first and then use AI tools to free up time for other learning objectives.
In den vergangenen Jahrzehnten hat es unübersehbar zahlreiche Fortschritte im Bereich der IT-Sicherheitsforschung gegeben, etwa in den Bereichen Systemsicherheit und Kryptographie. Es ist jedoch genauso unübersehbar, dass IT-Sicherheitsprobleme im Alltag der Menschen fortbestehen. Mutmaßlich liegt dies an der Komplexität von Alltagssituationen, in denen Sicherheitsmechanismen und Gerätefunktionalität sowie deren Heterogenität in schwer antizipierbarer Weise mit menschlichem Verständnis und Alltagsgebrauch interagieren. Um die wissenschaftliche Forschung besser auf Menschen und deren IT-Sicherheitsbedürfnisse auszurichten, müssen wir daher den Alltag der Menschen besser verstehen. Das Verständnis von Alltag ist in der Informatik jedoch noch unterentwickelt. Dieser Beitrag möchte das Forschungsfeld “Sicherheit in der Digitalisierung des Alltags” definieren, um Forschenden die Gelegenheit zu geben, ihre Anstrengungen in diesem Bereich zu bündeln. Wir machen dabei Vorschläge einerseits zur inhaltlichen Eingrenzung der informatischen Forschung. Andererseits möchten wir durch die Einbeziehung von Forschungsmethoden aus der Ethnografie, die Erkenntnisse aus der durchaus subjektiven Beobachtung des “Alltags” vieler einzelner Individuen zieht, zur methodischen Weiterentwicklung interdisziplinärer Forschung in diesem Feld beitragen. Die IT- Sicherheitsforschung kann dann Bestehendes gezielt für eine richtige Alltagstauglichkeit optimieren und neue grundlegende Sicherheitsfunktionalitäten für die konkreten Herausforderungen im Alltag entwickeln.
Understanding of financial data has always been a point of interest for market participants to make better informed decisions. Recently, different cutting edge technologies have been addressed in the Financial Technology (FinTech) domain, including numeracy understanding, opinion mining and financial ocument processing.
In this thesis, we are interested in analyzing the arguments of financial experts with the goal of supporting investment decisions. Although various business studies confirm the crucial role of argumentation in financial communications, no work has addressed this problem as a computational argumentation task. In other words, the automatic analysis of arguments. In this regard, this thesis presents contributions in the three essential axes of theory, data, and evaluation to fill the gap between argument mining and financial text.
First, we propose a method for determining the structure of the arguments stated by company representatives during the public announcement of their quarterly results and future estimations through earnings conference calls. The proposed scheme is derived from argumentation theory at the micro-structure level of discourse. We further conducted the corresponding annotation study and published the first financial dataset annotated with arguments: FinArg.
Moreover, we investigate the question of evaluating the quality of arguments in this financial genre of text. To tackle this challenge, we suggest using two levels of quality metrics, considering both the Natural Language Processing (NLP) literature of argument quality assessment and the financial era peculiarities.
Hence, we have also enriched the FinArg data with our quality dimensions to produce the FinArgQuality dataset.
In terms of evaluation, we validate the principle of ensemble learning on the argument identification and argument unit classification tasks. We show that combining a traditional machine learning model along with a deep learning one, via an integration model (stacking), improves the overall performance, especially in small dataset settings.
In addition, despite the fact that argument mining is mainly a domain dependent task, to this date, the number of studies that tackle the generalization of argument mining models is still relatively small. Therefore, using our stacking approach and in comparison to the transfer learning model of DistilBert, we address and analyze three real-world scenarios concerning the model robustness over completely unseen domains and unseen topics.
Furthermore, with the aim of the automatic assessment of argument strength, we have investigated and compared different (refined) versions of Bert-based models that incorporate external knowledge in the decision layer. Consequently, our method outperforms the baseline model by 13 ± 2% in terms of F1-score through integrating Bert with encoded categorical features.
Beyond our theoretical and methodological proposals, our model of argument quality assessment, annotated corpora, and evaluation approaches are publicly available, and can serve as strong baselines for future work in both FinNLP and computational argumentation domains.
Hence, directly exploiting this thesis, we proposed to the community, a new task/challenge related to the analysis of financial arguments: FinArg-1, within the framework of the NTCIR-17 conference.
We also used our proposals to react to the Touché challenge at the CLEF 2021 conference. Our contribution was selected among the «Best of Labs».
In the Internet of Things (IoT), Low-Power Wide-Area Networks (LPWANs) are designed to provide low energy consumption while maintaining a long communications’ range for End Devices (EDs). LoRa is a communication protocol that can cover a wide range with low energy consumption. To evaluate the efficiency of the LoRa Wide-Area Network (LoRaWAN), three criteria can be considered, namely, the Packet Delivery Rate (PDR), Energy Consumption (EC), and coverage area. A set of transmission parameters have to be configured to establish a communication link. These parameters can affect the data rate, noise resistance, receiver sensitivity, and EC. The Adaptive Data Rate (ADR) algorithm is a mechanism to configure the transmission parameters of EDs aiming to improve the PDR. Therefore, we introduce a new algorithm using the Multi-Armed Bandit (MAB) technique, to configure the EDs’ transmission parameters in a centralized manner on the Network Server (NS) side, while improving the EC, too. The performance of the proposed algorithm, the Low-Power Multi-Armed Bandit (LP-MAB), is evaluated through simulation results and is compared with other approaches in different scenarios. The simulation results indicate that the LP-MAB’s EC outperforms other algorithms while maintaining a relatively high PDR in various circumstances.
After the enactment of the GDPR in 2018, many companies were forced to rethink their privacy management in order to comply with the new legal framework. These changes mostly affect the Controller to achieve GDPR-compliant privacy policies and management.However, measures to give users a better understanding of privacy, which is essential to generate legitimate interest in the Controller, are often skipped. We recommend addressing this issue by the usage of privacy preference languages, whereas users define rules regarding their preferences for privacy handling. In the literature, preference languages only work with their corresponding privacy language, which limits their applicability. In this paper, we propose the ConTra preference language, which we envision to support users during privacy policy negotiation while meeting current technical and legal requirements. Therefore, ConTra preferences are defined showing its expressiveness, extensibility, and applicability in resource-limited IoT scenarios. In addition, we introduce a generic approach which provides privacy language compatibility for unified preference matching.
Network communication has become a part of everyday life, and the interconnection among devices and people will increase even more in the future. A new area where this development is on the rise is the field of connected vehicles. It is especially useful for automated vehicles in order to connect the vehicles with other road users or cloud services. In particular for the latter it is beneficial to establish a mobile network connection, as it is already widely used and no additional infrastructure is needed. With the use of network communication, certain requirements come along.
One of them is the reliability of the connection. Certain Quality of Service (QoS) parameters need to be met. In case of degraded QoS, according to the SAE level specification, a downgrade of the automated system can be required, which may lead to a takeover maneuver, in which control is returned back to the driver. Since such a handover takes time, prediction is necessary to forecast the network quality for the next few seconds. Prediction of QoS parameters, especially in terms of Throughput (TP) and Latency (LA), is still a challenging task, as the wireless transmission properties of a moving mobile network connection are undergoing fluctuation. In this thesis, a new approach for prediction Network Quality Parameters (NQPs) on Transmission Control Protocol (TCP) level is presented. It combines the knowledge of the environment with the low level parameters of the mobile network. The aim of this work is to perform a comprehensive study of various models including both Location Smoothing (LS) grid maps and Learning Based (LB) regression ones. Moreover, the possibility of using the location independence of a model as well as suitability for automated driving is evaluated.
The autonomic composition of Virtual Networks (VNs) and Service Function Chains (SFCs)based on application requirements is significant for complex environments. In this paper, we use graph transformation in order to compose an Extended Virtual Network (EVN) that is based on
different requirements, such as locations, low latency, redundancy, and security functions. The EVN can represent physical environment devices and virtual application and network functions. We build
a generic Virtual Network Embedding (VNE) framework for transforming an Application Request (AR) to an EVN. Subsequently, we define a set of transformations that reflect preliminary topological, performance, reliability, and security policies. These transformations update the entities and demands of the VN and add SFCs that include the required Virtual Network Functions (VNFs). Additionally, we propose a greedy proactive heuristic for path-independent embedding of the composed SFCs. This heuristic is appropriate for real complex environments, such as industrial networks. Furthermore, we present an Industrail Internet of Things (IIoT) use case that was inspired by Industry 4.0 concepts,in which EVNs for remote asset management are deployed over three levels; manufacturing halls and edge and cloud computing. We also implement the developed methods in Alevin and show exemplary mapping results from our use case. Finally, we evaluate the chain embedding heuristic while using a random topology that is typical for such a use case, and show that it can improve the admission ratio and resource utilization with minimal overhead.
The power demand (kW) and energy consumption (kWh) of data centers were augmenteddrastically due to the increased communication and computation needs of IT services. Leveragingdemand and energy management within data centers is a necessity. Thanks to the automated ICTinfrastructure empowered by the IoT technology, such types of management are becoming more feasiblethan ever. In this paper, we look at management from two different perspectives: (1) minimization of theoverall energy consumption and (2) reduction of peak power demand during demand-response periods.Both perspectives have a positive impact on total cost of ownership for data centers. We exhaustivelyreviewed the potential mechanisms in data centers that provided flexibilities together with flexiblecontracts such as green service level and supply-demand agreements. We extended state-of-the-artby introducing the methodological building blocks and foundations of management systems for theabove mentioned two perspectives. We validated our results by conducting experiments on a lab-gradescale cloud computing data center at the premises of HPE in Milano. The obtained results support thetheoretical model, by highlighting the excellent potential of flexible service level agreements in Green IT:33% of overall energy savings and 50% of power demand reduction during demand-response periods inthe case of data center federation.
Natural Language Processing has an important role in Artificial Intelligence for easing human-machine interaction. Processing human language, though, poses many challenges, among which is the semantics-related phenomenon known as language variability, the fact that the same thing can be said in several ways. NLP applications' inputs and outputs can be expressed in different forms, whose equivalence can be verified through inference. The textual entailment paradigm was established to enable the creation of a unifying framework for applied inference, providing a means of delivering other NLP task from handling inference issues in an ad-hoc manner, using instead the outputs of an inference-dedicated mechanism.
Text entailment, the task of determining whether a piece of text logically follows from another piece of text, involves different scenarios, which can range from a simple syntactic variation to more complex semantic relationships between sentences. However, most approaches try a one-size-fits-all solution that usually favors some scenario to the detriment of another. The commonsense world knowledge necessary to support more complex inferences is also usually employed in a limited way, with most approaches sticking to shallow semantic information, leaving more elaborate semantic relationships aside. Furthermore, most systems still work as a "black box", providing a yes/no answer that does not explain the underlying reasoning process.
This thesis aims at addressing these issues by proposing a composite interpretable approach for recognizing text entailment where the entailment pair is analyzed so the most relevant phenomenon is detected and the suitable method can be used to solve it. Syntactic variations are dealt with through the analysis of the sentences' syntactic structures, and semantic relationships are detected with the aid of a knowledge graph built from natural language dictionary definitions. Also, if a semantic matching is involved, the answer is made interpretable through the generation of natural language justifications that explain the semantic relationship between the pieces of text. The result is the XTE - Explainable Text Entailment - a system that outperforms well-established tools based on single-technique entailment algorithms, and that also gives an important step towards Explainable AI, allowing the inference model interpretation, making the semantic reasoning process explicit and understandable.
Programming is a key skill in a world where businesses are driven by digital transformations. Although many of the programming demand can be addressed by a simple set of instructions composing libraries and services available in the web, non-technical professionals, such as domain experts and analysts, are still unable to construct their own programs due to the intrinsic complexity of coding. Among other types of end-user development, natural language programming has emerged to allow users to program without the formalism of traditional programming languages, where a tailored semantic parser can translate a natural language utterance to a formal command representation able to be processed by a computational machine. Currently, semantic parsers are typically built on the top of a learning method that defines its behaviours based on the patterns behind a large training data, whose production frequently are costly and time-consuming. Our research is devoted to study and propose a semantic parser for natural language commands targeting a scenario with low availability of training data. Our proposed semantic parser follows a multi-component architecture, composed of a specialised shallow parser that associates natural language commands to predicate-argument structures, integrated to a distributional ranking model that matches the command to a function signature available from an API knowledge base. Systems developed with statistical learning models and complex linguistics resources, as the proposed semantic parser, do not provide natively an easy way to associate a single feature from the input data to the impact in system behaviour. In this scenario, end-user explanations for intelligent systems has become a strong requirement to increase user confidence and system literacy. Thus, our research designed an explanation model for the proposed semantic parser that fits the heterogeneity of its multi-component architecture. The explanation model explores a hierarchical representation in an increasing degree of technical depth, providing higher-level explanations in the initial layers, going gradually to those that demand technical knowledge, applying different explanation strategies to better express the approach behind each component. With the support of a user-centred experiment, we compared the utility of different types of explanations and the impact of background knowledge in their preferences.
In this thesis we consider real analytic functions, i.e. functions which can be described locally as convergent power series and ask the following: Which real analytic functions definable in R_{an,exp} have a holomorphic extension which is again definable in R_{an,exp}? Finding a holomorphic extension is of course not difficult simply by power series expansion. The difficulty is to construct it in a definably way.
We will not answer the question above completely, but introduce a large non trivial class of definable functions in R_{an,exp} where for example functions which are iterated compositions from either side of globally subanalytic functions and the global logarithm are contained. We call them restricted log-exp-analytic. After giving some preliminary results like preparation theorems and Tamm's Theorem for this class of functions we are able to show that real analytic restricted log-exp-analytic functions have a holomorphic extension which is again restricted log-exp-analytic.
Network virtualization provides high flexibility for deploying communication services in dense and heterogeneous environments. Two main approaches (dimensions) that are usually combined exist: Network Function Virtualization (NFV) technologies for functionality virtualization and Virtual Network Embedding (VNE) algorithms for resource virtualization. These approaches can be applied to different network levels, such as factory and enterprise levels of industrial networks. Several objectives and constraints, that might be conflicting, shall be considered when network virtualization is applied, mainly in complex topologies. This thesis proposes a network virtualization model that considers both virtualization dimensions, two network levels, and different objectives and constraints. The network levels considered are two primary levels in industrial networks. However, this consideration does not restrict the model to a particular environment or certain levels. The considered objectivities/constraints are topology, reliability, security, performance, and resource usage.
Based on this model, we first build an overall combined solution for autonomic and composite virtual networking. This solution considers both virtualization dimensions, two network levels, and target objectives. Furthermore, this solution combines three novel virtualization sub-approaches that consider performance, reliability, and performance. However, the sub-approaches apply to different combinations of levels and dimensions, and the reliability approach additionally considers the resource usage objective. After presenting all solutions, we map them to the defined model.
Regarding applicability to industrial networks, the combined approach is applied to an enterprise-level Industrial Internet of Things (IIoT) use case inspired by the smart factory concept in Industry 4.0. However, the sub-approaches are applied to more specific use cases. The performance and reliability solutions are integrated with relevant components of the Time Sensitive Networks (TSN) standard as a modern technology for industrial networks. The goal is to enrich the reliability and performance capabilities of TSN with the flexibility of network virtualization.
In the combined approach, we compose and embed an environment-aware Extended Virtual Network (EVN) that represents the physical devices, virtual application functions, and required Service Function Chains (SFCs). We use the graph transformation method to transform abstract application requirements (represented by an Application Request (AR)) into an EVN. Both EVN composition and embedding methods consider the Substrate Network (SN) topology and different security, reliability, performance, and resource usage policies. These policies are applied with a certain priority and depend on the properties of communicating entities such as location and type. The EVN is embedded using property-based node mapping, reliability-aware branching, and a greedy chain embedding heuristic. The chain embedding heuristic is evaluated using a random topology that represents the use case.
The performance sub-approach is NFV-based and is applied to a specific use case with Time-critical Traffic (TCT) flows. We develop and evaluate a complete framework for virtualizing Time-aware Shaper (TAS) using high-performance NFV. The reliability sub-approach is VNE-based and is applied to a specific factory level use case. We develop minimal and maximal branching heuristics based on a reliability-aware k-shortest path algorithm and compare them using a typical factory topology. We then integrate these algorithms with a Frame Replication and Elimination for Reliability (FRER) simulator to realize reliability policies by the autonomic and efficient configuration of a supporting technology.
The security sub-approaches are related to both virtualization dimensions and are applied to generic enterprise-level use cases. However, the applicability of the security aspect to industrial networks is only shown in the combined (EVN) approach and its use case. We research the autonomic security management in Network Function Virtualization Infrastructure (NFVI) with the main goal of early reaction to threats through SFC reconfiguration through Virtual Network Function (VNF) live migration. This goal is approached by supporting the security measurements with a decision making architecture that considers, on the one hand, the threats and events in the environment and, on the other hand, the Service Level Agreement (SLA) between the NFVI provider and user. For this purpose, we classify the VNF-specific attacks and define possible early detectable behavior patterns. Finally, we develop a security-aware VNE heuristic that considers the security requirements of the Virtual Network (VN) and the security capabilities of the SN. This approach is modified in the combined approach to consider deploying virtualized security VNFs.
Sentences that present a complex linguistic structure act as a major stumbling block for Natural Language Processing (NLP) applications whose predictive quality deteriorates with sentence length and complexity. The task of Text Simplification (TS) may remedy this situation. It aims to modify sentences in order to make them easier to process, using a set of rewriting operations, such as reordering, deletion or splitting. These transformations are executed with the objective of converting the input into a simplified output, while preserving its main idea and keeping it grammatically sound. State-of-the-art syntactic TS approaches suffer from two major drawbacks: first, they follow a very conservative approach in that they tend to retain the input rather than transforming it, and second, they ignore the cohesive nature of texts, where context spread across clauses or sentences is needed to infer the true meaning of a statement. To address these problems, we present a discourse-aware TS framework that is able to split and rephrase complex English sentences within the semantic context in which they occur. By generating a fine-grained output with a simple canonical structure that is easy to analyze by downstream applications, we tackle the first issue. For this purpose, we decompose a source sentence into smaller units by using a linguistically grounded transformation stage. The result is a set of selfcontained propositions, with each of them presenting a minimal semantic unit. To address the second concern, we suggest not only to split the input into isolated sentences, but to also incorporate the semantic context in the form of hierarchical structures and semantic relationships between the split propositions. In that way, we generate a semantic hierarchy of minimal propositions that benefits downstream Open Information Extraction (IE) tasks. To function well, the TS approach that we propose requires syntactically well-formed input sentences. It targets generalpurpose texts in English, such as newswire or Wikipedia articles, which commonly contain a high proportion of complex assertions.
In a second step, we present a method that allows state-of-the-art Open IE systems to leverage the semantic hierarchy of simplified sentences created by our discourseaware TS approach in constructing a lightweight semantic representation of complex assertions in the form of semantically typed predicate-argument structures. In that way, important contextual information of the extracted relations is preserved that allows for a proper interpretation of the output. Thus, we address the problem of extracting incomplete, uninformative or incoherent relational tuples that is commonly to be observed in existing Open IE approaches. Moreover, assuming that shorter sentences with a more regular structure are easier to process, the extraction of relational tuples is facilitated, leading to a higher coverage and accuracy of the extracted relations when operating on the simplified sentences. Aside from taking advantage of the semantic hierarchy of minimal propositions in existing Open IE Abstract approaches, we also develop an Open IE reference system, Graphene. It implements a relation extraction pattern upon the simplified sentences.
The framework we propose is evaluated within our reference TS implementation DisSim. In a comparative analysis, we demonstrate that our approach outperforms the state of the art in structural TS both in an automatic and a manual analysis. It obtains the highest score on three simplification datasets from two different domains with regard to SAMSA (0.67, 0.57, 0.54), a recently proposed metric targeted at automatically measuring the syntactic complexity of sentences which highly correlates with human judgments on structural simplicity and grammaticality. These findings are supported by the ratings from the human evaluation, which indicate that our baseline implementation DisSim returns fine-grained simplified sentences that achieve a high level of syntactic correctness and largely preserve the meaning of the input. Furthermore, a comparative analysis with the annotations contained in the RST Discourse Treebank (RST-DT) reveals that we are able to capture the contextual hierarchy between the split sentences with a precision of approximately 90% and reach an average precision of almost 70% for the classification of the rhetorical relations that hold between them. Finally, an extrinsic evaluation shows that when applying our TS framework as a pre-processing step, the performance of state-ofthe-art Open IE systems can be improved by up to 32% in precision and 30% in recall of the extracted relational tuples.
Accordingly, we can conclude that our proposed discourse-aware TS approach succeeds in transforming sentences that present a complex linguistic structure into a sequence of simplified sentences that are to a large extent grammatically correct, represent atomic semantic units and preserve the meaning of the input. Moreover, the evaluation provides sufficient evidence that our framework is able to establish a semantic hierarchy between the split sentences, generating a fine-grained representation of complex assertions in the form of hierarchically ordered and semantically interconnected propositions. Finally, we demonstrate that state-of-the-art Open IE systems benefit from using our TS approach as a pre-processing step by increasing both the accuracy and coverage of the extracted relational tuples for the majority of the Open IE approaches under consideration. In addition, we outline that the semantic hierarchy of simplified sentences can be leveraged to enrich the output of existing Open IE systems with additional meta information, thus transforming the shallow semantic representation of state-of-the-art approaches into a canonical context-preserving representation of relational tuples.
The increasing relevance of massive graph data reinforces the need for adequate graph data management. While several graph database engines have been developed, the storage of graph data in a relational database management system, and therefore the seamless integration into existing information systems remains an open challenge.
Motivated by the use case to integrate Building Information Modeling (BIM) data into the MonArch system, we propose a solution that transforms the BIM data into a property graph and stores this graph in the database system.
We present a novel approach to efficiently store property graph data in a relational database management system using JSON functionality and redundant storage of edges in adjacency lists and show how to import huge data sets into this schema. Applying this approach, we import data sets of up to nearly 1 TB of disk space within the relational database, while only having 96 GB of main memory available.
We also present a new approach of how to retrieve data from this database schema, translating queries written in the popular property graph query language Cypher into SQL. Hence, we provide an intuitive way to write semantically complex queries.
We also demonstrate the efficiency of our approach using the standardized Linked Data Benchmark Council – Social Network Benchmark (LDBC - SNB) framework. Our approach increases the throughput for this benchmark by up to 85 times, compared to existing approaches for RDBMS.
In addition, we propose a new method to transform BIM data into the property graph model and how to apply the aforementioned property graph storage to this data. We can import IFC models of up to 300 MB within five minutes.
We show the suitability of our approach using our own use case specific benchmark, which we integrated into the previously mentioned Social Network Benchmark. For our interactive use case-specific queries, we achieve response times faster than 5 ms in 99% of all executions.
Finally, we present how the aforementioned approach to store BIM data in a relational database management system is integrated into the existing MonArch system by splitting the different functionalities of our approach into a microservice architecture.
Critical infrastructure and contemporary business organizations are experiencing an ongoing paradigm shift of business towards more collaboration and agility. On the one hand, this shift seeks to enhance business efficiency, coordinate large-scale distribution operations, and manage complex supply chains. But, on the other hand, it makes traditional security practices such as firewalls and other perimeter defenses insufficient. Therefore, concerns over risks like terrorism, crime, and business revenue loss increasingly impose the need for enhancing and managing security within the boundaries of these systems so that unwanted incidents (e.g., potential intrusions) can still be detected with higher probabilities. To this end, critical infrastructure organizations step up their efforts to investigate new possibilities for actively engaging in situational awareness practices to ensure a high level of persistent monitoring as well as on-site observation.
Compliance with security standards is necessary to ensure that organizations meet regulatory requirements mostly shaped by a set of best practices. Nevertheless, it does not necessarily result in a coherent security strategy that considers the different aims and practical constraints of each organization. In this regard, there is an increasingly growing demand for risk-based security management approaches that enable critical infrastructures to focus their efforts on mitigating the risks to which they are exposed. Broadly speaking, security management involves the identification, assessment, and evaluation of long-term (or overall) objectives and interests as well as the means of achieving them.
Due to the critical role of such systems, their decision-makers tend to enhance the system resilience against very unpleasant outcomes and severe consequences. That is, they seek to avoid decision options associated with likely extreme risks in the first place. Practically speaking, this risk attitude can significantly influence the decision-making process in such critical organizations. Towards incorporating the aversion to extreme risks into security management decisions, this thesis investigates thoroughly the capabilities of a recently emerged theory of games with payoffs that are probability distributions. Unlike traditional optimization techniques, this theory provides an alternative decision technique that is more robust to extreme risks and uncertainty. Furthermore, this thesis proposes a new method that gives a decision maker more control over the decision-making process through defining loss regions with different importance levels according to people's risk attitudes. In this way, the static decision analysis used in the distribution-valued games is transformed into a dynamic process to adapt to different subjective risk attitudes or account for future changes in the decision caused by a learning process or other changes in the context.
Throughout their different parts, this thesis shows how theoretical models, simulation, and risk assessment models can be combined into practical solutions. In this context, it deals with three facets of security management: allocating limited security resources, prioritizing security actions, and tweaking decision making. Finally, the author discusses experiences and limitations distilled from this research and from investigating the new theory of games, which can be taken into account in future approaches.
The segmentation of volumetric datasets, i.e., the partitioning of the data into disjoint sub-volumes with the goal to extract information about these regions,is a difficult problem and has been discussed in medical imaging for decades.
Due to the ever-increasing imaging capabilities, in particular in X-ray computed tomography (CT) or magnetic resonance imaging, segmentation in industrial applications also gains interest.
Especially in industrial applications the generated datasets increase in size.
Hence, most applications apply well-known techniques in a 2+1-dimensional manner,i.e., they apply image segmentation procedures on each slice separately and track the progress along the axis of the volume in which the slices are stacked on.
This discards the information on preceding or subsequent slices, which is often assumed to be nearly identical. However, in the industrial context this might prove wrong since industrial parts might change their appearance significantly over the course of even a few slices.
Moreover, artifacts can further distort the content of the slices.
Therefore, three-dimensional processing of voxel volumes has to be preferred, which induces constraints upon the segmentation procedures. For example, they must not consider global information as it is usually not feasible in big scans to compute them efficiently.
Yet another frequent problem is that applications focus on individual parts only and algorithms are tailored to that case. Most prominent medical segmentation procedures do so by applying methods to specifically find the liver and only the liver of a patient, for example.
The implication is that the same method then cannot be applied to find other parts of the scan and such methods have to be designed individually for any object to be segmented.
Flexible segmentation methods are needed too specifically when partitioning unique scans. We define a unique scan to be a voxel dataset for which no comparable volume exists.
Classical examples include the use case of cultural heritage where not only the objects themselves are unique but also scan parameters are optimized to obtain the best image quality possible for that specific scan.
This thesis aims at introducing novel methods for voxelwise classifications based on local geometric features.
The latter are computed from local environments around each voxel and extract information in similar ways as humans do, namely by observing their similarity to geometric or textural primitives.
These features serve as the foundation to learning the proposed voxelwise classifiers and to discriminate between segmented and unsegmented voxels.
On the one hand, they perform fully automated clustering of volumes for which a representative random sample is extracted first.
On the other hand, a set of segmenting classifiers can be trained from few seed voxels, i.e., volume elements for which a domain expert marked if they belong to the components that shall be segmented. The interactive selection offers the advantage that no completely labeled voxel volumes are necessary and hence that unique scans of objects can be segmented for which no comparable scans exist.
Overall, it will be shown that all proposed segmentation methods are effectively of linear runtime with respect to the number of voxels in the volume. Thus, voxel volumes without size restrictions can be segmented in an efficient linear pass through the volume.
Finally, the segmentation performance is evaluated on selected datasets which shows that the introduced methods can achieve good results on scans from a broad variety of domains for both small and big voxel volumes.