How Toxic Can You Get? Search-based Toxicity Testing for Large Language Models
Corbo, S.; Bancale, L.; Gennaro, V. D.; Lestingi, L.; Scotti, V.; Camilli, M.
2025. IEEE Transactions on Software Engineering.
Research at KASTEL SRL
Research at KASTEL Security Research Labs is distinguished by its broad scope. Privacy and reliability aspects are naturally part of IT security; basic and applied research complement each other, and a comprehensive approach to IT security needs to look not just at the technical system, but also at its social environment as well as its human aspects.
-
Fact 1
666 Publications from 2021 to 2025
-
Fact 2
148 Papers at 42 Top Conferences 2021 to 2025
-
Fact 3
65 Awards from 2021 to 2025
Our research covers a broad spectrum of cybersecurity. It combines the view of different disciplines in order to adress not only technical aspects, but also human factors and the legal framework.
Research Projects
Security for the Internet of Medical Things
Together with Wibu-Systems AG, KASTEL Security Research Labs is researching the project “Security for the Internet of Medical Things”, or “Sec4IoMT” for short. The aim is to find and exemplarily implement a solution approach for the long-term preservation of the security of Internet of Medical Things (IoMT) devices in critical infrastructures. When updating the security of individual components, the interoperability of the devices should be maintained.. At the same time, a security update must not lead to a loss of certification for the medical devices.
Research Projects
Secure Federated Machine Learning
Machine learning models require large amounts of data to be trained on, leaving little training data available in use cases with strong privacy requirements.The Secure Federated Machine Learning project of SAP and KASTEL Security Research Labs is investigating approaches for training a neural network on data from multiple parties without having to reveal their data to the other parties. These approaches can then be used in use cases with strong privacy requirements.
KASTEL Security Research Labs conduct interdisciplinary and application-oriented research. Therefore, we permanently maintain and expand numerous partnerships with industrial enterprises and research institutions. The resulting synergy effects flow into our research.
-
Fact 1
156 Academic Cooperations in 25 Countries
-
Fact 2
24 Ongoing Thirt-Party Funded Projects
-
Fact 3
92 Cooperations with Companies
Cooperations and Exchanges
Propolis Privacy for Smart Cities
2021, the German-French cooperation PROPOLIS was launched. Prof. Thorsten Strufe from KASTEL Security Research Labs at KIT, together with SAP, the Urban Software Institute GmbH, Chemnitz, and EURECOM Frankreich, is analyzing AI based data protection for smart cities and citizens. The project will be concluded in 2024. Read more about PROPOLIS
Cooperations and Exchanges
SofDCar – Software-Defined Car
Since August 2021, a total of 13 companies and research institutions have been working in the publicly funded “Software-Defined Car” project (SofDCar for short). The experts from industry and science want to create standardized rules and processes together for the car of the future and its effective use of data. The project is publicly funded by the German Federal Ministry of Economic Affairs and Climate Action (BMWK). Read more about SofDCar
Publications
Offene und subkutane Zielkonflikte zwischen KI-VO und DSGVO – Das Zusammenspiel zwischen Vertrauenswürdigkeit von KI und Vertrauen in Datenverarbeitung als grundlegende Herausforderung
Winau, M.; Spiecker genannt Döhmann, I.
2025. Computer und Recht : CR ; Zeitschrift für die Praxis des Rechts der Informationstechnologien, (10), 647–654
Proactive self-adaptation and assurance of explainable Human–Machine Teaming
Lestingi, L.; Bersani, M. M.; Camilli, M.; Mirandola, R.; Rossi, M.; Scandurra, P.
2026. Journal of Systems and Software, 232, Art.-Nr.: 112657.
BFId: Identity Inference Attacks Utilizing Beamforming Feedback Information
Todt, J.; Morsbach, F.; Strufe, T.
2025. Proceedings of 32nd ACM SIGSAC Conference on Computer and Communications Security (CCS ’25), Taipei, October 13–17, 2025, Association for Computing Machinery (ACM).
NOTHORG; A digital twin for ECU security testing
Wehmer, M.; Diestl, J.; Waibel, D.; Baumgart, I.
2025. Proceedings of the 2025 Cyber Security in CarS Workshop (CSCS ’25), 9 S., Association for Computing Machinery (ACM).
Towards the Derivation of Encryption Rules for Encryption-Based Access Control for Delta-Based Model Representations — The DEBAC Approach
Weber, T.; Quellmalz, K.; Benz, L.; Weber, S.; Burger, E.; Reussner, R.
2025. Second International Workshop on Model Management (MoM), co-located with the 28th International Conference on Model Driven Engineering Languages and Systems (MODELS 2025), Grand Rapids, MI, 5th-10th October 2025, Institute of Electrical and Electronics Engineers (IEEE)