ICT Dissertation Award

Scientific achievements and the further qualification of employees are the core responsibility of the individual Fraunhofer institutes. The Fraunhofer ICT Group supports these goals by awarding internal prizes for outstanding publications and theses at the Fraunhofer Institutes.

Since 2015, the Fraunhofer ICT Group has presented the annual Fraunhofer ICT Dissertation Award. Here, the jury awards outstanding dissertations from the Fraunhofer institutes that deal with highly innovative developments and technologies in computer science, mathematics or related fields. The prize is endowed with 5,000 euros, 3,000 euros and 2,000 euros for the first three places.

1^st prize

Fraunhofer FIT – Rezaul Karim

His work "Interpreting Black-Box Machine Learning Models with Decision Rules and Knowledge Graph Reasoning" aims to improve the interpretability and explainability of non-transparent machine learning models without significantly compromising prediction accuracy. To this end, various representation learning techniques are used to transform multimodal data into a common latent space. Interpretable methods are then applied to generate decision rules and negative factors that explain how the model works. Robustness against attacks and the integration of symbolic logic are also considered. Quantitative evaluation has shown that the approach achieves higher performance compared to existing methods.

2^nd prize

Fraunhofer AISEC – Franziska Boenisch

The thesis "Secure and Private Machine Learning" examines security and privacy aspects in the field of machine learning (ML). It is shown that ML practitioners have a low awareness of data protection and rely on third-party frameworks and services. New attack vectors and security vulnerabilities are highlighted and the need for secure and privacy-compliant design of ML methods is emphasized.

3^rd prize

Fraunhofer IDMT – Stylianos Ioannis Mimilakis

The thesis "Deep Learning-Based Music Source Separation" investigates the problem of music source separation using deep learning methods. Three aspects are considered: signal processing, neural architecture and signal representation. New algorithms and network architectures are presented that show that deep neural networks can learn data-driven filters for music source separation. The proposed methods have achieved promising results in music source separation.

The 2023 award winners come from the institutes Fraunhofer IGD, Fraunhofer ITWM and Fraunhofer SIT.  In our interview they explain which topics their work deals with and why they are so interesting on the one hand, but at the same time so important for our future.

1^st prize

Fraunhofer ITWM – Sebastian Blauth

Dr. Sebastian Blauth of the Fraunhofer Institute for Industrial Mathematics ITWM won the award for his paper titled "Adjoint-Based Shape Optimization and Optimal Control with Applications to Microchannel Systems."

2^nd Prize

Fraunhofer IGD – Philipp Terhörst

Dr.-Ing. Philipp Terhörst from the Fraunhofer Institute for Computer Graphics Research IGD was awarded second place for his work. His dissertation is entitled "Mitigating Soft-Biometric Driven Bias and Privacy Concerns in Face Recognition Systems".

3^rd Prize

Fraunhofer SIT – Andreas Fuchs

The third prize went to Dr.-Ing. Andreas Fuchs from the Fraunhofer Institute for Secure Information Technology SIT and his work on "Engineering Trustworthy Systems by Minimizing and Strengthening their TCBs using Trusted Computing".

1^st Prize

Fraunhofer IEM – Johannes Späth

This year's ICT Dissertation Award went to Johannes Späth from the Fraunhofer Institute for Mechatronics Design IEM, who passed his doctoral exam with the highest grade "summa cum laude". He completed his doctorate at the University of Paderborn on the topic of "Synchronized Pushdown Systems for Pointer and Data-Flow Analysis" under Prof. Eric Bodden.  

His dissertation involves the development of efficient and accurate data-flow analysis algorithms. The algorithms help software developers to detect errors in their program code early, during compilation. The algorithms achieve efficiency through "on demand" analysis; instead of analyzing the entire software, only relevant parts of the software code are analyzed. Thus, results are delivered within seconds, instead of hours as with existing analysis algorithms. At the same time, the algorithms are highly precise with an extremely low false positive rate.

1^st Prize

Fraunhofer IOSB – Julius Pfrommer

»Distributed Planning for Self-Organizing Production Systems«

The work describes an automatic adaptation of the control of production plants to changing orders and framework conditions. The principle of self-organization through distributed planning is applied.

The evaluation of the thesis with the rating "with distinction", the numerous publications of the author and the high citation rates (h-index: 13) underline the high scientific level of the work. At the same time, the work is of very high practical relevance for the manufacturing industry and thus has a strong economic and thus also social impact.

2^nd Prize

Fraunhofer MEVIS – André Homeyer

»Automated analysis of necrosis and steatosis in histological images – Practical solutions for coping with heterogeneity and variability«

The dissertation makes important contributions to the improvement of automated histological image analysis with a focus on the quantification of necrosis or the quantification of steatosis in histological sections of liver tissue. Newly developed scores enable reliable measurement of heterogeneously distributed tissue properties. Interactive training leads to reduced effort while still providing accurate results. Finally, methods are presented that enable accurate yet rapid automated analysis of even large image data sets.

Due to the thematic focus, the work has a very high societal relevance and also exhibits a high degree of interdisciplinarity, innovation as well as transferability and generalizability.

3rd Prize

Fraunhofer AISEC – Konstantin Böttinger

»Fuzzing with Stochastic Feedback Processes«

In the context of the thesis, new AI-supported testing methods are presented to support modern software development processes and thus to improve the security of software. For this purpose, results from probability theory are translated into algorithms for testing software.

The work provides relevant new methods to secure complex software systems against cyber attacks. This aspect will become immensely important in view of the advancing digitalization and networking and the increasing importance of software in the industrial, governmental and private sectors. Since potentially vulnerable software is used to control critical infrastructures, industrial plants, production processes, medical devices and automated driving vehicles, practical approaches to solutions for increasing software and IT security have an enormous economic and social impact. In view of the fundamental and still increasing importance of IUK technologies for society, high attention should be paid to cognitive security.

A unique twist characterizes the awarding of prizes in 2019: For the first time, two first prizes and one third prize were awarded.

The first prize is shared by Dr. Anna-Marie Kruspe from the Fraunhofer Institute for Digital Media Technology IDMT in Ilmenau and Dr. Erion Elmasllari from the Fraunhofer Institute for Applied Information Technology FIT in Sankt Augustin.

1^st Prize

Fraunhofer IDMT – Anna-Marie Kruspe

Anna-Marie Kruspe was awarded for her work "Application of Automatic Speech Recognition Technologies to Singing". The field of Music Information Retrieval deals with the automatic analysis of musical characteristics. One aspect that has hardly been explored so far is the sung lyrics. In automatic speech recognition, many methods are developed for the automatic analysis of speech, but rarely for singing. The work of Anna-Marie Kruspe investigates the application of methods from speech recognition to singing and describes ways to apply and improve automatic speech recognition techniques to various singing-related search tasks, including phoneme recognition, speech recognition, keyword search, text-to-song matching, and song text search. To this end, she pursued two general approaches: First, she works on better phoneme recognition models, which she identified as a general bottleneck in almost all of the above retrieval tasks, and trains them with enhanced speech recordings presented in a more "song-like" manner, or with real song recordings with automatically matched lyrics. Second, she adapts automatic speech recognition techniques to song, incorporating domain knowledge and adapting them to the various properties of this data.

1^st Prize

Fraunhofer FIT – Erion Elmasllari

Erion Elmasllari's dissertation, "A Framework for the Successful Design and Deployment of Electronic Triage Systems," was convincing. Triage is the process by which emergency responders, especially in high patient volume situations, determine the order and priority for medical assistance. After assessing the severity of injuries, first responders tag victims to indicate a priority for treatment. At the end of the triage process, victims are treated according to the assigned priority.

Erion Elmasllari's work identifies the key factors that determine the acceptance of e-triage systems and uses them as the basis for a framework for designing such systems. The framework can be used evaluatively to communicate the reasons why previous e-Triage systems have failed, but can also be used generatively to actively assist system designers in designing acceptable e-Triage systems. In either case, the framework illuminates the needs, limitations, risks, and issues of e-Triage systems within the overarching triage ecosystem.

3^rd Prize

Fraunhofer IGD – Naser Damer

The third prize was awarded to Dr. Naser Damer from the Fraunhofer Institute for Computer Graphics Research IGD in Darmstadt for the doctoral thesis entitled "Application-driven Advances in Multi-biometric Fusion". Biometrics is the automated recognition of individuals based on their behavior or biological characteristics. Typical applications include fingerprint or facial image recognition to gain access to IT systems or entry to rooms. The major advantage of biometrics is that the user does not have to remember a password or carry a key. To create the most secure system possible, numerous approaches have been proposed that seek a balance between, for example, security, accuracy, and ease of use. One relatively new and promising approach is multibiometric fusion. Here, multiple biometric features are used to identify a particular user. Naser Damer's work focuses on such multibiometric fusion systems with the goal of improving overall performance and functionality. By optimizing the fusion process, his approaches achieve higher biometric accuracy through the use of additional information. He also presents specific applications that complement biometric systems and use multibiometric fusion to improve performance.

Once again this year, the Fraunhofer Institute for Secure Information Technology SIT can claim first place in the Fraunhofer ICT Dissertation Awards.

1^st Prize

Fraunhofer SIT – Steven Arzt

Dr. Steven Arzt's work on "Static Data Flow Analysis for Android Applications" was convincing across the board. In it, he presents the FlowDroid tool, which extracts precise and largely complete data flows from popular apps such as Facebook, PayPal and LinkedIn. FlowDroid's result reports improve the user's digital sovereignty, as they can now see for themselves how an app handles their data before installing the app on their device and thus entrusting it with their data. With FlowDroid, the user can understand what data is collected from his device and sent to third parties.

2^nd Prize

Fraunhofer IOSB – Philipp Christoph Sebastian Bier

In his work "Implementation of the Right to Information under Data Protection Law on the Basis of Usage Control and Data Provenance Technologies," Dr. Philipp Christoph Sebastian Bier subjects the right to information in data protection law to a critical appraisal and creates comprehensive technical requirements for exercising this right to information. He designs and implements a distributed, data-centric, integrated and data subject-focused data protection information system with which the right to information can be exercised interactively and step-by-step. On the one hand, this requires in-depth legal preparation of the problem and, on the other, equally in-depth IT penetration. This is the only way to arrive at a solution that is both scientifically and technically state of the art, but at the same time meets the highest legal requirements. Dr. Bier has succeeded very well in finding such a solution. This work from the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB was worth the second prize to the jury.

3^rd Prize

Fraunhofer ITWM – Sebastian Osterroth

This year's third prize went to the Fraunhofer Institute for Industrial Mathematics ITWM with the work of Dr. Sebastian Osterroth on "Mathematical models for the simulation of combined depth and cake filtration processes". In this paper, Dr. Osterroth describes the modeling of filtration processes. In this process, solids of different particle sizes accumulate from the substance flowing through the filter, which on the one hand improve the filtration process, but on the other hand increase the flow resistance. The resulting pressure drop essentially determines the energy required for the filtration process. In the dissertation, a combination of deep-bed filtration and cake filtration is considered. In deep-bed filtration, the separation of particles occurs in the flowed filter medium. Particle separation is achieved solely by the filter membrane placed in the flow. In cake filtration, on the other hand, it is desired that filtration be improved by the filter cake that forms as a result of deposition. Dr. Osterroth developed a new mathematical model for the coupled process consisting of a nonlinear system of partial and ordinary differential equations.

On the evening of September 26, 2017, this year's Fraunhofer ICT Dissertation Award winners were selected in Chemnitz. The award ceremony traditionally took place during the annual conference of the Gesellschaft für Informatik.

1^st Prize

Fraunhofer SIT – Siegfried Rasthofer

In the third edition of the award, which is endowed with a total of EUR 10,000, Dr. Siegfried Rasthofer from Fraunhofer SIT in Darmstadt won first prize with his work "Improving Mobile-Malware Investigations with Static and Dynamic Code Analysis Techniques".
In his work, he developed an approach to analyze Android software for malicious content. In particular, this solution can be used to check third-party apps for malicious content, even when no source code is available. Malicious code can be detected even when developers attempt to bypass this detection. The results of the dissertation have given rise to the product "CodeInspect" at Fraunhofer SIT, which has been licensed by various manufacturers for malicious code detection.

2^nd Prize

Fraunhofer MEVIS – Federico von Samson-Himmelstjerna

The second prize was awarded to Dr. Federico von Samson-Himmelstjerna for his work "Robust and time-efficient determination of perfusion parameters using time-encoded arterial spin labeling MRI", which he wrote at Fraunhofer MEVIS in Bremen. Behind this is a medical procedure for measuring blood flow using magnetic resonance imaging. The special feature here is that the patient does not have to be administered a contrast medium for this measurement. The procedure can be used to efficiently support the diagnosis and monitoring of diseases such as stroke, multiple sclerosis, Alzheimer's, Parkinson's or cancer of the brain.

3^rd Prize

Fraunhofer IDMT – Alexander Loos

Dr. Alexander Loos from Fraunhofer IDMT in Ilmenau dealt with a not quite everyday application of image analysis and face recognition technologies in his doctoral thesis "Face Recognition for Great Apes: Identification of Primates in Real-World Environments", for which he received the third prize of the Fraunhofer ICT Dissertation Award. His method represents the world's first real-time capable system for the individual recognition of great apes in images and videos, which also works reliably under difficult open-field conditions. It is used in wildlife research to evaluate conservation measures for the great apes, which are threatened with extinction. The method is equally suitable for use in zoos and wildlife parks to provide visitors with a more interactive experience.

The ICT Dissertation Awards were presented for the second time on September 28, 2016 as part of the Computer Science Day of the German Informatics Society in Klagenfurt. The winning PhD students of this year's award came from the institutes Fraunhofer ITWM, Fraunhofer IGD and Fraunhofer FIT.

1^st Prize

Fraunhofer ITWM – Jan Kleinert

Dr. rer. nat. Jan Kleinert received the 1st prize with his work "Simulating Granular Material using Nonsmooth Time-Stepping and a Matrix-Free Interior Point Method", TU Kaiserslautern, submitted by Prof. Dr. Dieter Prätzel-Wolters of Fraunhofer ITWM.

2^nd Prize

Fraunhofer IGD – Tobias Große-Puppendahl

Dr.-Ing. Tobias Große-Puppendahl received the 2nd prize with his work "Capacitive Sensing and Communication for Ubiquitous Interaction and Environmental Perception", TU Darmstadt, submitted by Prof. Dr. Dieter W. Fellner from Fraunhofer IGD.

3^rd Prize

Fraunhofer FIT – Katja Niemann

The third prize went to Dr. rer. nat. Katja Niemann for the paper "Discovery of Usage-based Item Similarities to Support Recommender Systems in Dealing with Rarely Used Items", RWTH Aachen, submitted by Prof. Dr. Matthias Jarke from Fraunhofer FIT.

1^st Prize

Fraunhofer MEVIS – Christian Rieder

Dr. Christian Rieder was awarded the inaugural "Fraunhofer ICT Dissertation Award" for his doctoral thesis "Interactive Visualization for Assistance of Needle-Based Interventions". The award was presented on September 30, 2015 during the 45th annual conference of the Gesellschaft für Informatik in Cottbus.

The award-winning work deals with a highly innovative visualization methodology that makes it easier to plan minimally invasive interventions for the treatment of tumors and to determine the success of the therapy more objectively.