You can also find my publications on Google Scholar or Semantic Scholar, however, they may not be up to date.
Tobias Lorenz, Marta Kwiatkowska, Mario Fritz in German Conference on Pattern Recognition (GCPR), 2024
We propose FullCert, the first end-to-end certifier with sound, deterministic robustness guarantees against both training- and inference-time attacks.
Alaa Anani, Tobias Lorenz, Bernt Schiele, Mario Fritz in International Conference on Machine Learning (ICML), 2024
This paper introduces, for the first time, an adaptive hierarchical certification algorithm for image semantic segmentation that certifies image pixels within a hierarchy and proves the correctness of its guarantees.
Tobias Lorenz, Marta Kwiatkowska, Mario Fritz in 16th ACM Workshop on Artificial Intelligence and Security (AISec), 2023
We propose the first systematic analysis of availability attacks against neural network certifiers and demonstrate highly effective backdoor attacks that exploit the fallback strategy. Initial defenses were unable to mitigate this novel attack.
Tobias Lorenz, Anian Ruoss, Mislav Balunović, Gagandeep Singh and Martin Vechev in IEEE/CVF International Conference on Computer Vision (ICCV), 2021
We introduce 3DCertify, the first verifier able to certify robustness of point cloud models to semantic transformations.
Yongqin Xian, Tobias Lorenz, Bernt Schiele and Zeynep Akata in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018
We propose a novel generative adversarial network that synthesizes CNN features conditioned on class-level semantic information, offering a shortcut directly from a semantic descriptor of a class to a class-conditional feature distribution.