Publications - Chair of Computer Science II

Full Papers

Oratio Homomorphico: Creating a Lecture for Homomorphic Encryption by applying Design-Based Research. Prantl, Thomas; Schneider, Tobias; Horn, Lukas; Engel, Simon; Iffländer, Lukas; Krupitzer, Christian; Bonilla, Rafael; Kounev, Samuel; in Proceedings of the 4th International Conference on Computer Science and Education in Computer Science (CSECS 2025), Bratislava, Slovakia, T. Zlateva, G. Tuparov (eds.), (2027). 31–45. Springer Nature Switzerland, Cham.
- [ Abstract ]
- [ BibTeX ]
- [ DOI ]
The ever-increasing pace of digitalization, with its collection of data and the training of artificial intelligence on this data, is bringing the topic of privacy-preserving computing more and more into focus. One promising technology for implementing the privacy-preserving computing paradigm is homomorphic encryption, which has been driven forward in recent years and still holds great potential for improvement in the future. The training of future computer scientists in this field is therefore of crucial importance in order to put this key technology into practice in the future. However, to the best of the authors' knowledge, there is currently no course that provides students with insights into the theoretical foundations of homomorphic encryption and teaches the practical applicability and extensibility of existing homomorphic cryptosystems. To close this gap, we present our own lecture in this paper, which focuses on the combination of theory and application of homomorphic encryption. We created our lecture material according to the Design-Based Research principle. We held and evaluated our lecture at three German universities. Our lecture evaluations have shown that we have been able to create suitable lecture material on homomorphic encryption and its application, which we would like to share with other lecturers.

@inproceedings{Oratio_Sardinen_2025, abstract = {The ever-increasing pace of digitalization, with its collection of data and the training of artificial intelligence on this data, is bringing the topic of privacy-preserving computing more and more into focus. One promising technology for implementing the privacy-preserving computing paradigm is homomorphic encryption, which has been driven forward in recent years and still holds great potential for improvement in the future. The training of future computer scientists in this field is therefore of crucial importance in order to put this key technology into practice in the future. However, to the best of the authors' knowledge, there is currently no course that provides students with insights into the theoretical foundations of homomorphic encryption and teaches the practical applicability and extensibility of existing homomorphic cryptosystems. To close this gap, we present our own lecture in this paper, which focuses on the combination of theory and application of homomorphic encryption. We created our lecture material according to the Design-Based Research principle. We held and evaluated our lecture at three German universities. Our lecture evaluations have shown that we have been able to create suitable lecture material on homomorphic encryption and its application, which we would like to share with other lecturers.}, address = {Cham}, author = {Prantl, Thomas and Schneider, Tobias and Horn, Lukas and Engel, Simon and Iffländer, Lukas and Krupitzer, Christian and Bonilla, Rafael and Kounev, Samuel}, booktitle = {Proceedings of the 4th International Conference on Computer Science and Education in Computer Science (CSECS 2025), Bratislava, Slovakia}, editor = {Zlateva, Tanya and Tuparov, Georgi}, keywords = {t_interdisciplinary}, month = {07}, pages = {31–45}, publisher = {Springer Nature Switzerland}, series = {EAI CSECS 2025}, title = {Oratio Homomorphico: Creating a Lecture for Homomorphic Encryption by applying Design-Based Research}, year = 2027 }
Privacy-Preserving Training and Inference of CNNs: A Medical Performance Case Study. Prantl, Thomas; Schneider, Louis; Engel, Simon; Horn, Lukas; Borst, Vanessa; Krupitzer, Christian; Kounev, Samuel; Bonilla, Rafael; in Proceedings of the 5th Intelligent Cybersecurity Conference (2025). 303–311. IEEE.
- [ BibTeX ]
- [ DOI ]
@inproceedings{MPC_Sardinen_2025, author = {Prantl, Thomas and Schneider, Louis and Engel, Simon and Horn, Lukas and Borst, Vanessa and Krupitzer, Christian and Kounev, Samuel and Bonilla, Rafael}, booktitle = {Proceedings of the 5th Intelligent Cybersecurity Conference}, keywords = {descartes}, month = {05}, pages = {303-311}, publisher = {IEEE}, series = {ICSC2025}, title = {Privacy-Preserving Training and Inference of CNNs: A Medical Performance Case Study}, year = 2025 }
Security Analysis of a Decentralized, Revocable and Verifiable Attribute-Based Encryption Scheme. Prantl, Thomas; Lauer, Marco; Horn, Lukas; Engel, Simon; Dingel, David; Kounev, Samuel; Bauer, André; Krupitzer, Christian; Kounev, Samuel; in Proceedings of the 19th International Conference on Availability, Reliability and Security (2024). 1–11. ACM.
- [ BibTeX ]
- [ DOI ]
@inproceedings{ARES_2024_ABE, author = {Prantl, Thomas and Lauer, Marco and Horn, Lukas and Engel, Simon and Dingel, David and Kounev, Samuel and Bauer, André and Krupitzer, Christian and Kounev, Samuel}, booktitle = {Proceedings of the 19th International Conference on Availability, Reliability and Security}, keywords = {prantl_spezial}, month = {07}, pages = {1-11}, publisher = {ACM}, series = {ARES 2024}, title = {Security Analysis of a Decentralized, Revocable and Verifiable Attribute-Based Encryption Scheme}, year = 2024 }

Journal and Magazine Articles

SoK: Poisoning Attacks and Defenses in Federated and Privacy-Preserving Machine Learning. Prantl, Thomas; Kulle, Edda; Horn, Lukas; Engel, Simon; Kounev, Samuel; in Proceedings on Privacy Enhancing Technologies (2027). 2027 Privacy Enhancing Technologies Symposium Advisory Board.

To appear.
- [ BibTeX ]
@article{Sardine_2027_Pets_Poison_SoK, author = {Prantl, Thomas and Kulle, Edda and Horn, Lukas and Engel, Simon and Kounev, Samuel}, journal = {Proceedings on Privacy Enhancing Technologies}, keywords = {myown}, month = {jul}, note = {To appear.}, publisher = {Privacy Enhancing Technologies Symposium Advisory Board}, title = {SoK: Poisoning Attacks and Defenses in Federated and Privacy-Preserving Machine Learning}, volume = 2027, year = 2027 }
Homomorphic approximation of non-linear activation functions using the CKKS cryptosystem of the OpenFHE library. Engel, Simon; Prantl, Thomas; Horn, Lukas; Bonilla, Rafael; Kounev, Samuel; in Discover Artificial Intelligence (2026). 6(1) Springer Science and Business Media LLC.
- [ BibTeX ]
- [ DOI ]
@article{engel2026homomorphic, author = {Engel, Simon and Prantl, Thomas and Horn, Lukas and Bonilla, Rafael and Kounev, Samuel}, journal = {Discover Artificial Intelligence}, keywords = {t_journalmagazine}, month = {aug}, number = 1, publisher = {Springer Science and Business Media LLC}, title = {Homomorphic approximation of non-linear activation functions using the CKKS cryptosystem of the OpenFHE library}, volume = 6, year = 2026 }
Performance analysis of the homomorphic implementation of K-Means using CKKS. Prantl, Thomas; Amann, Patrick; Horn, Lukas; Engel, Simon; Bauer, Andr{é}; Bonilla, Rafael; Benhnini, Anwar; Kounev, Samuel; in Discover Applied Sciences (2026). 8(7) 744.

https://doi.org/10.1007/s42452-026-09088-w
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
- [ DOI ]
Modern data science requires large amounts of data and computational power to perform effectively. However, there is a lot of sensitive data that should not be shared with others. To still do data science, methods like Homomorphic Encryption exist that enable calculations on encrypted data. One well-known data science algorithm is K-Means, which clusters data points into a previously fixed number of clusters. We provided a homomorphic implementation by demonstrating how comparisons and assignments of clusters can be calculated. Furthermore, the algorithm is evaluated in an offline approach with bootstrapping and an online approach with client re-encryption. Both achieve an accuracy of 100{\%}, but take significantly longer, ranging from 50,000 to 16 million times compared to the non-encrypted case.

@article{Sardine_k-Means_1, abstract = {Modern data science requires large amounts of data and computational power to perform effectively. However, there is a lot of sensitive data that should not be shared with others. To still do data science, methods like Homomorphic Encryption exist that enable calculations on encrypted data. One well-known data science algorithm is K-Means, which clusters data points into a previously fixed number of clusters. We provided a homomorphic implementation by demonstrating how comparisons and assignments of clusters can be calculated. Furthermore, the algorithm is evaluated in an offline approach with bootstrapping and an online approach with client re-encryption. Both achieve an accuracy of 100{\%}, but take significantly longer, ranging from 50,000 to 16 million times compared to the non-encrypted case.}, author = {Prantl, Thomas and Amann, Patrick and Horn, Lukas and Engel, Simon and Bauer, Andr{\'e} and Bonilla, Rafael and Benhnini, Anwar and Kounev, Samuel}, journal = {Discover Applied Sciences}, keywords = {myown}, month = {jul}, note = {https://doi.org/10.1007/s42452-026-09088-w}, number = 7, pages = 744, title = {Performance analysis of the homomorphic implementation of K-Means using CKKS}, volume = 8, year = 2026 }
Quo Vadis CKKS: Comparison of the Realization of Basic Mathematical Functions for the Homomorphic Cryptosystem CKKS using De Bello and Polynomial Approximations. Prantl, Thomas; Horn, Lukas; Engel, Simon; Bauer, André; Kounev, Samuel; in Journal of Information Security and Applications (2025). 93 Elsevier.

Published online June 2025
- [ BibTeX ]
- [ URL ]
- [ DOI ]
@article{HE_Sardinen_2025, author = {Prantl, Thomas and Horn, Lukas and Engel, Simon and Bauer, André and Kounev, Samuel}, journal = {Journal of Information Security and Applications}, keywords = {prantl_spezial}, month = {sep}, note = {Published online June 2025}, publisher = {Elsevier}, title = {Quo Vadis CKKS: Comparison of the Realization of Basic Mathematical Functions for the Homomorphic Cryptosystem CKKS using De Bello and Polynomial Approximations}, volume = 93, year = 2025 }
De Bello Homomorphico: Investigation of the extensibility of the OpenFHE library with basic mathematical functions by means of common approaches using the example of the CKKS cryptosystem. Prantl, Thomas; Horn, Lukas; Engel, Simon; Iffländer, Lukas; Beierlieb, Lukas; Krupitzer, Christian; Bauer, André; Sakarvadia, Mansi; Foster, Ian; Kounev, Samuel; in International Journal of Information Security (2024). 23(2) 1149–1169.

Published online November 2023
- [ BibTeX ]
- [ DOI ]
@article{PrHoEnIfBeKrBaSaFoKo_IJoIS_SGC_Pareto, author = {Prantl, Thomas and Horn, Lukas and Engel, Simon and Iffl\"ander, Lukas and Beierlieb, Lukas and Krupitzer, Christian and Bauer, Andr\'e and Sakarvadia, Mansi and Foster, Ian and Kounev, Samuel}, journal = {International Journal of Information Security}, keywords = {prantl_spezial}, month = 4, note = {Published online November 2023}, number = 2, pages = {1149–1169}, title = {De Bello Homomorphico: Investigation of the extensibility of the OpenFHE library with basic mathematical functions by means of common approaches using the example of the CKKS cryptosystem}, volume = 23, year = 2024 }
Towards a Cryptography Encyclopedia: A Survey on Attribute-Based Encryption. Prantl, Thomas; Zeck, Timo; Horn, Lukas; Iffländer, Lukas; Bauer, André; Dmitrienko, Alexandra; Krupitzer, Christian; Kounev, Samuel; in Journal of Surveillance, Security and Safety (2023).
- [ BibTeX ]
- [ DOI ]
@article{PrZeHoIfBaDmKrKo_ABESurvey, author = {Prantl, Thomas and Zeck, Timo and Horn, Lukas and Iffl\"ander, Lukas and Bauer, Andr\'e and Dmitrienko, Alexandra and Krupitzer, Christian and Kounev, Samuel}, journal = {Journal of Surveillance, Security and Safety}, keywords = {myown}, title = {Towards a Cryptography Encyclopedia: A Survey on Attribute-Based Encryption}, year = 2023 }
Benchmarking of Secure Group Communication schemes with Focus on IoT. Prantl, Thomas; Bauer, André; Engel, Simon; Horn, Lukas; Krupitzer, Christian; Iffl{ä}nder, Lukas; Kounev, Samuel; in Discover Data (2024). 2(1) Springer Science and Business Media LLC.
- [ BibTeX ]
- [ URL ]
- [ DOI ]
@article{PrBaEnHoKrIfKo_Discover_Data_2024, author = {Prantl, Thomas and Bauer, André and Engel, Simon and Horn, Lukas and Krupitzer, Christian and Iffl{\"a}nder, Lukas and Kounev, Samuel}, journal = {Discover Data}, keywords = {myown}, month = {may}, number = 1, publisher = {Springer Science and Business Media LLC}, title = {Benchmarking of Secure Group Communication schemes with Focus on IoT}, volume = 2, year = 2024 }

Books

Laying the Foundation for Homomorphic Training of Neural Networks Horn, Lukas Johannes; in BestMasters (2026). Springer Vieweg Wiesbaden.

To appear September 2026
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
As Artificial Intelligence becomes increasingly integrated into modern technology, the tension between data utility and individual privacy intensifies. Privacy-Preserving Machine Learning (PPML) addresses this conflict by utilizing techniques such as Homomorphic Encryption (HE), which enables computation on encrypted data. However, implementing HE in neural networks necessitates the use of functional approximations for complex non-linear operations, introducing errors that may affect model performance. This book investigates the impact of these approximations through a novel simulation framework. By evaluating various network architectures—including Multi-Layer Perceptrons and Convolutional Neural Networks—across multiple datasets, the author analyses the effects of Taylor series, Newton-Raphson, and Chebyshev polynomial approximations, alongside realistic Gaussian noise. The findings reveal that neural networks are inherently resilient to these errors, often maintaining or even improving baseline accuracy. By establishing the necessary precision levels for common activation functions, this work demonstrates the feasibility of efficient homomorphic training, marking a vital advancement for secure and private machine learning.

@book{HornFoundationHomomorphicTraining, abstract = {As Artificial Intelligence becomes increasingly integrated into modern technology, the tension between data utility and individual privacy intensifies. Privacy-Preserving Machine Learning (PPML) addresses this conflict by utilizing techniques such as Homomorphic Encryption (HE), which enables computation on encrypted data. However, implementing HE in neural networks necessitates the use of functional approximations for complex non-linear operations, introducing errors that may affect model performance. This book investigates the impact of these approximations through a novel simulation framework. By evaluating various network architectures—including Multi-Layer Perceptrons and Convolutional Neural Networks—across multiple datasets, the author analyses the effects of Taylor series, Newton-Raphson, and Chebyshev polynomial approximations, alongside realistic Gaussian noise. The findings reveal that neural networks are inherently resilient to these errors, often maintaining or even improving baseline accuracy. By establishing the necessary precision levels for common activation functions, this work demonstrates the feasibility of efficient homomorphic training, marking a vital advancement for secure and private machine learning.}, author = {Horn, Lukas Johannes}, keywords = {t_book}, note = {To appear September 2026}, publisher = {Springer Vieweg Wiesbaden}, series = {{{BestMasters}}}, title = {Laying the {{Foundation}} for {{Homomorphic Training}} of {{Neural Networks}}}, year = 2026 }

Workshop Papers

Cooking in the Dark: Performance Evaluation of Matrix-Vector Operations in Neural Networks under Homomorphic Encryption. Horn, Lukas; Svoboda, Lennart; Engel, Simon; Prantl, Thomas; Kounev, Samuel; in SSP 2026 (2026).

To appear.
- [ BibTeX ]
- [ Download ]
@inproceedings{HoSvEnPrKo-CookingDark2026, author = {Horn, Lukas and Svoboda, Lennart and Engel, Simon and Prantl, Thomas and Kounev, Samuel}, booktitle = {SSP 2026}, keywords = {t_workshop}, note = {To appear.}, title = {Cooking in the Dark: Performance Evaluation of Matrix-Vector Operations in Neural Networks under Homomorphic Encryption}, year = 2026 }
Classification in the Dark: Performance Insights and Novel Methods for Homomorphic Binary Networks. Horn, Lukas; Svoboda, Lennart; Engel, Simon; Prantl, Thomas; Kounev, Samuel; in WAHC 2026 (2026).

To appear.
- [ BibTeX ]
@inproceedings{HoSvEnPrKo-ClassifyingDark2026, author = {Horn, Lukas and Svoboda, Lennart and Engel, Simon and Prantl, Thomas and Kounev, Samuel}, booktitle = {WAHC 2026}, keywords = {t_workshop}, note = {To appear.}, title = {Classification in the Dark: Performance Insights and Novel Methods for Homomorphic Binary Networks}, year = 2026 }

Short Papers

Performance Impact Analysis of Homomorphic Encryption: A Case Study Using Linear Regression as an Example. Prantl, Thomas; Engel, Simon; Horn, Lukas; Kaiser, Dennis; Iffländer, Lukas; Bauer, André; Krupitzer, Christian; Kounev, Samuel; in Information Security Practice and Experience, W. Meng, Z. Yan, V. Piuri (eds.), (2023). 284–298. Springer Nature Singapore, Singapore.
- [ Abstract ]
- [ BibTeX ]
- [ DOI ]
In recent years, the trend has increasingly been to store and process data in the cloud. However, this is based on the premise that cloud providers treat the data in a trustworthy manner. One way of using the data in the cloud without the provider having access to it is homomorphic encryption. However, since this encryption has only recently become practicable, analysis of its for practical applications is still in its infancy. Therefore, we investigate the performance of homomorphic encryption using a real-world application, namely linear regression. Our main finding is that although the homomorphic computation of linear regression is in the range of minutes and thus slower than in the non-homomorphic case, linear regression can be computed homomorphic and is therefore suitable for use cases where data security is the top priority.

@inproceedings{PrEnHoDeIfBaKrKo_ISPEC_De_Bello, abstract = {In recent years, the trend has increasingly been to store and process data in the cloud. However, this is based on the premise that cloud providers treat the data in a trustworthy manner. One way of using the data in the cloud without the provider having access to it is homomorphic encryption. However, since this encryption has only recently become practicable, analysis of its for practical applications is still in its infancy. Therefore, we investigate the performance of homomorphic encryption using a real-world application, namely linear regression. Our main finding is that although the homomorphic computation of linear regression is in the range of minutes and thus slower than in the non-homomorphic case, linear regression can be computed homomorphic and is therefore suitable for use cases where data security is the top priority.}, address = {Singapore}, author = {Prantl, Thomas and Engel, Simon and Horn, Lukas and Kaiser, Dennis and Iffl\"ander, Lukas and Bauer, Andr\'e and Krupitzer, Christian and Kounev, Samuel}, booktitle = {Information Security Practice and Experience}, editor = {Meng, Weizhi and Yan, Zheng and Piuri, Vincenzo}, keywords = {t_short}, pages = {284–298}, publisher = {Springer Nature Singapore}, title = {Performance Impact Analysis of Homomorphic Encryption: A Case Study Using Linear Regression as an Example}, year = 2023 }

Poster Papers

Towards Performance-Oriented Planning and Development of IoT Applications and Networks. Prantl, Thomas; Horn, Lukas; Engel, Simon; Meissner, Maximilian; Bonilla, Rafael; Kounev, Samuel; in Companion of the 17th ACM/SPEC International Conference on Performance Engineering (2026). 92–93. Association for Computing Machinery, Italy.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
- [ DOI ]
The planning of IoT applications requires consideration of the performance, energy consumption, and security of resource-constrained devices. In particular, n-to-n communication patterns and security mechanisms significantly affect the runtime, energy consumption, and reliability of IoT systems. In this article, we summarize our previous work on the performance and security analysis of secure group communication and attribute-based encryption methods, as well as on the evaluation of publish/subscribe brokers. Based on measurements on real-world IoT hardware, we outline an approach for modeling and combining individual IoT operations to estimate time and energy consumption in advance. The goal is to support developers in performance-oriented planning and development of energy-efficient, secure IoT applications.

@inproceedings{10.1145/3777911.3799728, abstract = {The planning of IoT applications requires consideration of the performance, energy consumption, and security of resource-constrained devices. In particular, n-to-n communication patterns and security mechanisms significantly affect the runtime, energy consumption, and reliability of IoT systems. In this article, we summarize our previous work on the performance and security analysis of secure group communication and attribute-based encryption methods, as well as on the evaluation of publish/subscribe brokers. Based on measurements on real-world IoT hardware, we outline an approach for modeling and combining individual IoT operations to estimate time and energy consumption in advance. The goal is to support developers in performance-oriented planning and development of energy-efficient, secure IoT applications.}, address = {New York, NY, USA}, author = {Prantl, Thomas and Horn, Lukas and Engel, Simon and Meissner, Maximilian and Bonilla, Rafael and Kounev, Samuel}, booktitle = {Companion of the 17th ACM/SPEC International Conference on Performance Engineering}, keywords = {myown}, pages = {92–93}, publisher = {Association for Computing Machinery}, series = {ICPE Companion '26}, title = {Towards Performance-Oriented Planning and Development of IoT Applications and Networks}, year = 2026 }

Bildnachweise