Qualifikationsarbeiten

Grundvorraussetzungen

Generell sind wir offen für Studierende aller Bereiche aus dem FB4, jedoch sollten Sie idealerweise mindestens eine der folgenden Veranstaltungen besucht und bestanden haben:

* Grundlagen der IT-Sicherheit,

* Datenschutz,

* Security in Networked Systems (SiNS),

* IT-Risk Management,

* Proseminar / Seminar in IT-Sicherheit,

* Projektpraktikum / Forschungspraktikum in IT-Sicherheit.

Vergabe von Qualifikationsarbeiten:

Hier finden Sie Informationen zur Vergabe von Abschlussarbeiten. Senden Sie bei Interesse bitte eine E-Mail an itsec-lehre@uni-koblenz.de mit

- Studiengang,

- aktuellem Notenauszug (KLIPS),

- dem Thema, das ausgeschrieben wurde,

- einem kurzen Motivationsschreiben,

- und dem von Ihnen gewünschten Bearbeitungszeitraum (von/bis),

oder bei eigenen Vorschlägen oder Wünschen

- eine Liste mit Themen, die Sie gerne bearbeiten würden.

Je konkreter Sie bei Ihrer Beschreibung für ein Wunschthema sind, desto schneller und besser können wir auf dieses eingehen. Idealerweise, jedoch nicht zwingend notwendig, reichen Sie nebst Thema ein kleines Exposé ein, in dem Sie den Hintergrund erläutern und ggf. eine Forschungsfrage (mit sicherheitsrelevanten Aspekt) definieren

Ablauf und Betreuung:

Vor der Anmeldung der Abschlussarbeit, unabhängig von möglichen Kollaborationen mit weiteren Parteien, einigt man sich mittels eines Proposals bzw. Exposés nochmal genau auf das Thema. Hierbei erarbeiten und formulieren Sie unter Betreuung eine Forschungsfrage.

Dieses Proposal umfasst 4-6 Seiten und besteht idealerweise aus den Kapiteln:

1. Einführung

2. Grundlagen / Hintergrund

3. Verwandte Arbeiten

4. Methodik

5. Zeitplan (vorzugsweise Gantt-Chart)

6. Limitationen / Ausblick (optional)

7. Literaturverzeichnis

Proposal:

Wir treffen uns bis zu zweimal zum Besprechen des Proposals.

Thesis:

Nach der Anmeldung der Arbeit wird erwartet, dass die Studierenden am monatlichen Oberseminar teilnehmen, bei dem individuelle Fragen zur Arbeit gestellt werden können.
Außerdem bieten wir ein monatliches Gespräch mit dem Betreuer an.
3-4 Monate nach Anmeldung sollte die Struktur und die Inhaltsangabe der Arbeit vollständig sein.
6-8 Wochen vor Abgabe der Arbeit sollte uns ein Entwurf der Arbeit zugesandt werden, den wir dann kommentieren (die Arbeit wird von den Betreuern nicht korrekturgelesen, allerdings werden wir Sie auf eventuelle gravierende Mängel hinweisen.)

Ausgeschriebene Arbeiten:

Im weiteren finden Sie hier die ausgeschriebenen Arbeiten und Kooperationsarbeiten. Beachten Sie, dass auch für diese die oben genannten Grundvoraussetzungen gelten.

Master Thesis: Knowledge Extraction from Function Plots

Art der Abschlussarbeit: Master (Bachelor)
Studiengang: Inf/ WInf/ CV/W&D Sci.

Kooperationsarbeit mit Partner: DIQA GmbH

Description

Datasheets for electronic components are documents that provide detailed information about a particular component, such as its physical characteristics, operating characteristics, and ratings. They also contain function plots that display important data describing certain properties of components, e.g. temperatures via voltage curves, or impedance via frequency curves. Function plots come in different levels of complexity, e.g. a curve or a set of curves, linear scale or logarithmic scale etc.

In this thesis, you will develop a method that automatically samples function plots of low complexity level and stores it in a datamodel.

As part of the large, nationwide research project progressivKI, the thesis will investigate how we can extract knowledge from documents in a text-based lessons learned repository using modern machine learning methods. ProgressivKI will develop a modularly structured AI platform to support the development of efficient and safe electronic systems for future vehicle applications (e.g. automated driving) https://www.edacentrum.de/projekte/progressivKI.

Goals of the Thesis

Get an overview of existing research in the subject area
Generate synthetic data consisting of function plots and numeric and textual information.
Identify promising neural network architectures that extract numeric and textual information from function plots.
Implement and evaluate identified architectures with synthetic and real-word data.

Basics

Python
Machine learning, neural networks, deep learning (pytorch)

Requirements

Basic knowledge of Python, git
Proactivity

The thesis requires a good knowledge of AI methods but also of application programming. Interessted students should apply via email including a short statement of interest and a copy of your transcript. Email to itsec-lehre@uni-koblenz.de.

About DIQA Projektmanagement GmbH

DIQA is a software company from Karlsruhe and offers solutions and services for Smart Enterprise Portals, especially search and knowledge management solutions. For this we use technologies from the Semantic Web as well as machine learning from texts.

https://www.diqa.de/

Master Thesis: Image Classification in datasheets

Art der Abschlussarbeit: Master (Bachelor)
Studiengang: Inf/ WInf/ CV/W&D Sci.
Kooperationsarbeit mit Partner: DIQA GmbH

Description

Datasheets for electronic components are documents that provide detailed information about a particular component, such as its physical characteristics, operating characteristics, and ratings. They also contain various types of images, like application circuits, schematics, function plots, package shots, block diagrams. Currently it is not possible for users to retrieve e.g., all impedance curves for an electronic component.
In this thesis, you are developing and evaluating neural network architectures to classify these images and assess the varacity of the solution. You will deploy a graphical annotation tool that enables end users to create datasets with classified images. You will use these datasets to evaluate your architectures. Thus, your work will greatly improve the daily work of end users.
As part of the large, nationwide research project progressivKI, the thesis will investigate how we can extract knowledge from documents in a text-based lessons learned repository using modern machine learning methods. ProgressivKI will develop a modularly structured AI platform to support the development of efficient and safe electronic systems for future vehicle applications (e.g. automated driving) https://www.edacentrum.de/projekte/progressivKI.

Goals of the Thesis

Get an overview of existing works on the subject
Create datasets consisting of images and numeric and textual information.
Identify promising neural network architectures that classify these images.
Implement and evaluate promising architectures with realword data.

Basics

Python
Machine learning, neural networks, deep learning (pytorch)

Requirements

Basic knowledge of Python, git
Proactivity

About DIQA Projektmanagement GmbH

https://www.diqa.de/

Master Thesis: IoT (Fraunhofer FKIE)

Art der Abschlussarbeit: Master (Bachelor)
Studiengang: Inf/ WInf/ CV/W&D Sci.
Kooperationsarbeit mit Fraunhofer FKIE

Introduction

The Internet of Things (IoT) is finding more and more applications in the public domain, and the number of open data sources keeps growing. This allows anyone to collect and analyse information from these sources. The information can be used for various purposes, such as traffic analysis, environmental monitoring, infrastructure protection, and even defence. But how can we verify that this data is trustworthy and has not been corrupted? This thesis investigates how to build trust in public IoT data and establish resilience against targeted misinformation by the information source and falsification by third parties, and what role artificial intelligence (AI) in particular can play in this regard.

Background

Using public IoT data for decision making and planning requires a high level of confidence in the integrity and accuracy of that data, especially in the context of a military operation. Incorrect or falsified information can lead to wrong decisions being made. Therefore, decision-relevant information is a high-value target for adversaries. Accordingly, it is important to develop mechanisms to build trust in public IoT data and to increase resilience against falsification. In this context, artificial intelligence offers promising potential to increase trust in public IoT data. By using AI-based algorithms, automated procedures can be developed to detect suspicious activity, identify anomalies, and verify the authenticity of IoT data. The use of AI can thus enable more effective detection and mitigation of tampering attempts to ensure the reliability and security of public IoT data in a military context.

Research Questions

What is the importance of trust in public IoT data for the acceptance and use of IoT applications in a civil/military context?
How can algorithms for detecting misinformation in public IoT data be developed and applied to ensure data integrity and trustworthiness? How can AI be used to enhance trust and detect misinformation?
What strategies and mechanisms can be used to verify the information source of public IoT data and effectively assess its trustworthiness?
What strategies and mechanisms can be used to increase the resilience of public IoT data against deliberate falsification, thereby increasing trust in data integrity?
What is the role of AI in these algorithms, policies, and mechanisms?
What best practices and recommendations, including AI, already exist to promote trust and improve the resilience of public IoT data in different civilian application domains (e.g., traffic analysis, environmental monitoring)?
How can identified (AI) mechanisms be implemented and evaluated in a laboratory demonstrator and what conclusions can be drawn?

Methodology

The work will focus on a comprehensive literature review first, looking at scientific papers and case studies to provide an overview of current methods for monitoring public IoT data for misinformation and falsification by third parties, and for increasing trust in this data. For this purpose, scientific literature as well as existing use cases will be reviewed and evaluated.

In a second step, a practical case study will be conducted, in which an AI-based solution for improving data integrity will be implemented and tested in a real IoT system. The focus will be on detecting anomalies in the collected data and investigating how well AI is able to detect and report such changes. For this purpose, a real IoT scenario will be designed and implemented in the lab using a demonstrator to investigate the benefits of AI with respect to the research questions.

Expected Results

As a result of this work, a comprehensive analysis of the existing challenges and vulnerabilities related to trust in public IoT data will be presented. For this purpose, a systematic review of existing literature and recent research results will be conducted, and the mechanisms described therein will be evaluated and compared. In addition, a laboratory demonstrator will be developed to test AI-based mechanisms for detecting anomalies in IoT data. The resulting findings will be analysed to derive a possible architecture for an AI-based detection system, but also to identify possible challenges and limitations in applying AI in IoT systems. Finally, the most important results of this thesis will be summarized, practical recommendations will be given, and future research directions will be pointed out.

Possible Paper Sources

· L. Wei, Y. Yang, J. Wu, C. Long and B. Li, "Trust Management for Internet of Things: A Comprehensive Study," in IEEE Internet of Things Journal, vol. 9, no. 10, pp. 7664-7679, 15 May15, 2022, doi: 10.1109/JIOT.2021.3139989.

Trust Management for Internet of Things: A Comprehensive Study | IEEE Journals & Magazine | IEEE Xplore

· A. Adewuyi, H. Cheng, Q. Shi, J. Cao, Á. MacDermott and X. Wang, "CTRUST: A Dynamic Trust Model for Collaborative Applications in the Internet of Things," in IEEE Internet of Things Journal, vol. 6, no. 3, pp. 5432-5445, June 2019, doi: 10.1109/JIOT.2019.2902022.

· CTRUST: A Dynamic Trust Model for Collaborative Applications in the Internet of Things | IEEE Journals & Magazine | IEEE Xplore Anomaly Detection in Internet of Things Using Artificial Intelligence: A Review" von R. Singh, P. Kumar, S. Tyagi, and S. Gupta

Attack-and-Anomaly-Detection-in-IoT-Networks-using-Machine-Learning-Techniques-A-Review.pdf (researchgate.net)

· A Trust Management Framework for the Internet of Things Using Artificial Intelligence" von S. Bhattacharya, A. Sengupta, and S. Roy

Sensors | Free Full-Text | A Trust Management Model for IoT Devices and Services Based on the Multi-Criteria Decision-Making Approach and Deep Long Short-Term Memory Technique (mdpi.com)

· Z. Lv, Y. Han, A. K. Singh, G. Manogaran and H. Lv, "Trustworthiness in Industrial IoT Systems Based on Artificial Intelligence," in IEEE Transactions on Industrial Informatics, vol. 17, no. 2, pp. 1496-1504, Feb. 2021, doi: 10.1109/TII.2020.2994747.

https://ieeexplore.ieee.org/abstract/document/9093957