Machine learning (ML) has been developed and applied mostly without considering aspects of information security. However, otherwise very well functioning models can be easily evaded by slight manipulations of an attacker—making it possible to exploit systems protected by ML. The study of these manipulations, known as adversarial examples (AEs), is a very active field of research that thrives for several years. Central questions about the origin and effects of AE are to be answered. As ML is applied to intrusion detection systems to protect IT infrastructure from unforeseen attacks with often fatal consequences, my work focuses on increasing their robustness with respect to AEs.

Additional interests of mine:

• machine learning in embedded systems
• free software

• Interpolation in the eXtended Classifier System: An architectural perspective. Anthony Stein, Dominik Rauh, Sven Tomforde, Jörg Hähner, Journal of Systems Architecture, Volume 75, 2017, Pages 79–94, ISSN 1383-7621. DOI: 10.1016/j.sysarc.2017.01.010
• Augmenting the Algorithmic Structure of XCS by Means of Interpolation. Stein A., Rauh D., Tomforde S., Hähner J., 2016 Architecture of Computing Systems (ARCS), Lecture Notes in Computer Science, vol 9637. Springer, Cham. DOI: 10.1007/978-3-319-30695-7_26
• Interpolation-based classifier generation in XCSF. A. Stein, C. Eymüller, D. Rauh, S. Tomforde and J. Hähner, 2016 IEEE Congress on Evolutionary Computation (CEC), Vancouver, BC, 2016, pp. 3990–3998. DOI: 10.1109/CEC.2016.7744296
• Dealing with Unforeseen Situations in the Context of Self-Adaptive Urban Traffic Control: How to Bridge the Gap. A. Stein, S. Tomforde, D. Rauh and J. Hähner, 2016 IEEE International Conference on Autonomic Computing (ICAC), Würzburg, 2016, pp. 167–172. DOI: 10.1109/ICAC.2016.20