h1

% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@PHDTHESIS{Welten:1003572,
      author       = {Welten, Sascha Martin},
      othercontributors = {Decker, Stefan Josef and Kirsten, Toralf},
      title        = {{M}ethods for practical data sharing and decentralised
                      analytics: an integrated approach},
      school       = {RWTH Aachen University},
      type         = {Dissertation},
      address      = {Aachen},
      publisher    = {RWTH Aachen University},
      reportid     = {RWTH-2025-01136},
      pages        = {1 Online-Ressource : Illustrationen},
      year         = {2025},
      note         = {Veröffentlicht auf dem Publikationsserver der RWTH Aachen
                      University; Dissertation, RWTH Aachen University, 2025},
      abstract     = {Sharing data across organisational borders and domains has
                      become indispensable for scientific progress and innovation,
                      mainly due to the growing trend of Artificial Intelligence
                      (AI) and data-driven approaches, which require large volumes
                      of data for optimal performance. However, concerns about
                      data privacy, security, and regulatory compliance often
                      hinder effective data sharing. As a result, data remains
                      siloed and largely inaccessible for research or industry
                      purposes. Although numerous data-sharing concepts have been
                      proposed, they often lack practical realisation and
                      evaluation in real-life scenarios. This dissertation
                      addresses these shortcomings by contributing an integrated
                      framework that combines privacy-preserving data-sharing
                      methods with Decentralised Analytics (DA). Based on current
                      and emerging data-sharing policies and regulations, this
                      dissertation first investigates and identifies several key
                      requirements essential for data sharing. These include
                      establishing trust, ensuring controlled data access,
                      managing distributed data sources, and addressing issues
                      related to data heterogeneity and utility. For each derived
                      requirement, this work conceptualises, implements, and
                      evaluates various proof of concepts, which are subsequently
                      integrated into a lifecycle called ’DAOps’. This
                      DevOps-inspired lifecycle offers a structured and seamless
                      workflow for managing data analysis processes on shared
                      data. To bring this lifecycle into operation and real-world
                      application, a novel data-sharing platform called
                      ’Platform for Analytics and Distributed Machine Learning
                      for Enterprises’ (PADME) is developed, which implements
                      the DAOps lifecycle. The last part of this dissertation
                      covers the evaluation of PADME in its entirety across five
                      research studies in healthcare and hydrology. The evaluation
                      confirms the applicability and operational readiness of
                      PADME in real-world research scenarios. The platform
                      supports a broad spectrum of analysis types, from basic
                      statistics to advanced Machine Learning (ML), while enabling
                      multi-institutional collaborations and managing data with
                      heterogeneous types and varying volumes. It adheres to
                      established data management standards and addresses the
                      increasing demand for ML applications by enabling
                      decentralised model training. The findings demonstrate that
                      the decentralised approaches achieve performance levels
                      comparable to those developed with traditional centralised
                      methods. This suggests that decentralised approaches offer a
                      viable and privacy-preserving alternative to conventional
                      data analysis techniques, which may lack privacy protection
                      and encounter regulatory challenges. Additionally, the
                      evaluation emphasises that common data schema standards and
                      well-balanced data distributions are critical drivers for
                      successful data sharing and DA. In conclusion, this
                      dissertation contributes to the emerging need for
                      data-sharing platforms and brings theoretical data-sharing
                      concepts into practice. The research demonstrates the
                      feasibility of data-driven research with DA through PADME
                      and provides insights into how automated methods support
                      systematic and secure data analysis. Ultimately, the
                      outcomes of this dissertation fuel research collaborations,
                      data-driven innovations, as well as privacy-preserving and
                      data-sovereign sharing of data between stakeholders.},
      cin          = {124510 / 120000},
      ddc          = {004},
      cid          = {$I:(DE-82)124510_20160614$ / $I:(DE-82)120000_20140620$},
      typ          = {PUB:(DE-HGF)11},
      doi          = {10.18154/RWTH-2025-01136},
      url          = {https://publications.rwth-aachen.de/record/1003572},
}