Offene Abschlussarbeiten

This thesis aims to investigate the reliability and trustworthiness of traffic sign classifiers when subjected to occlusions. Utilizing the German Traffic Sign Recognition Benchmark (GTSRB) dataset, this research will focus on annotating the dataset with various levels and types of occlusions to evaluate if the predictions are still trustworthy. The primary objective is to assess the performance degradation of the classifier under different occlusion scenarios and to develop strategies to enhance its robustness. This study is crucial for improving the safety and reliability of autonomous driving systems where traffic signs might be partially obscured.

The aim of this thesis is the analysis and prototypical evaluation of different replication strategies in which computations on highly volatile data structures are outsourced to different remote nodes. The thesis should explore the solution space in terms of consistency and latency properties, timeliness as well as migration capabilites. As a concrete example, the work should examine the scenario of an automotive application that replicates its local application state onto nearby multi-access edge computing nodes that will then run computationally heavy calculations.

This thesis will investigate methods for detecting natural adversarial examples against ImageNet classifiers using classic computer vision techniques. Adversarial examples are inputs to machine learning models that are designed to cause the model to make a mistake. This project will utilize the Harder ImageNet Test Set (https://arxiv.org/abs/1907.07174) as an dataset for Natural Adversarial Examples. The primary objective is to explore and compare the effectiveness of traditional computer vision methods, such as histograms and SIFT (Scale-Invariant Feature Transform), in identifying these adversarial examples. The outcome of this research will enhance our understanding of model vulnerabilities and contribute to developing more robust machine learning systems.

This thesis will explore the effectiveness of different optimization strategies in the generation of adversarial patches. Adversarial patches are small, intentionally designed perturbations that can cause machine learning models, particularly in computer vision, to misclassify inputs. The primary objective of this research is to compare various optimization techniques, such as gradient-based methods, evolutionary algorithms, and reinforcement learning, to determine which methods are most effective and efficient in creating these patches. The outcome of this research could significantly enhance our understanding of model vulnerabilities and contribute to the development of more robust machine learning systems.

Subjective Logic (SL) is a mathematical framework for reasoning under uncertainty. It is useful for expressing opinions on how reliable information is (so-called Trust Opinions) and performing computations on these opinions. At our institute, we research applications of SL in the automotive domain, e.g. to express trust in data received from a sensor or from other vehicles. Current implementations of SL internally use floating-point arithmetic (IEEE 754) for performing calculations. However, IEEE 754 floating-point numbers are prone to introducing rounding errors. In safety-critical domains, failing to account for such errors might lead to catastrophic consequences. In this thesis/project, you will investigate the potential impact of floating-point errors in SL calculations and develop strategies to minimise it. You can choose your approach freely: whether you work theoretically (e.g. through a detailed study of literature) or practically (e.g. through implementing a test environment and explaining the observed effects) is up to you.

Abstract: Policy Languages such as XACML or ALFA are well-known and well-defined in the area of access control. With Zero Trust Service Function Chaining (ZTSFC) [https://journal.ub.tu-berlin.de/eceasst/article/view/1138], an advanced Zero Trust (ZT) architecture, new requirements came up for such Policy Languages. The goal of the thesis is to set up a list of this requirements, to identify missing features in existing policy languages. Based on this, the most promising policy language is to be extended by this missing features.

An autonomous vehicle is equipped with a variety of sensors that produce large quantites of data which the vehicle uses to run a lot of different safety-critical functions, such as Cooperative Adaptive Cruise Control or Park Assist. In this thesis, we focus on the trust between the vehicle computer and other in-vehicle components that it relies upon to provide non-compromised data as input to different safety-critical functions. The goal of the thesis is to build a tool that will automate building of in-vehicular trust models based on a system model of a vehicle. A system model of a simplified vehicle will first need to be built by using the System Modeling Language (SysML). This model will serve as an input to the automation tool that needs to output a trust model in form of a Subjective Trust Network. The methodology for building such trust models within the framework of Subjective Logic will be provided.

Service Function Chaining (SFC) is a technice to steer traffic through specific network services. To proof that the traffic was actually forwarded through the specified services, a Proof Of Transit (PoT) is used. In this project, different PoT approaches are compared and the most promising solution implemented in a HTTPS-based SFC environment.