Fabio Brau

Email: fabio.brau@unica.it

Addr: Via Marengo, 2, 09123

City: Cagliari, Italy

Fabio holds a Master’s degree in Mathematics (Pisa, 2019) and a Ph.D. in Computer Engineering (Pisa, 2023), where his research focused on the certification of robustness for deep neural networks. His current work spans Adversarial Machine Learning and Trustworthiness of AI Systems, with focus on Computer Vision. He has also contributed to the theoretical and practical study of Lipschitz-constrained architectures for enhancing model stability and certifiable robustness. He published in top-tier AI/ML conferences such as NeurIPS, CVPR, AAAI and, in journals such as IEEE TPAMI. He is currently involved in European research projects on trustworthy AI, and serves as a reviewer for several conferences and journals.

News

Oct 29, 2026	Our paper Latent-space Attacks for Refusal Evasion in Language Models has been accepted at NeurIPS 2026! See you in Sydney! Huge congratulations to Giorgio Piras and Raffaele Mura!
Oct 29, 2026	Our paper Out-of-Distribution Detection for Continual Learning: Design Principles and Benchmarking has been accepted in the International Journal of Computer Vision! Huge congratulations to Srishti Gupta for leading the work!
Jul 24, 2026	Our paper Over-parameterization and Adversarial Robustness in Neural Networks has been accepted in Discover Artificial Intelligence! Huge congratulations to Srishti Gupta for leading the work!
Jun 03, 2026	New blog post: The Tick That Poisoned Your Training Set — alpha-gal syndrome read as a textbook data-poisoning backdoor attack on the immune system.

Latest Posts

Jun 02, 2026	The Tick That Poisoned Your Training Set

Selected Publications

NeurIPS

Latent-space Attacks for Refusal Evasion in Language Models

Giorgio Piras^*, Raffaele Mura^*, Fabio Brau^*, and 4 more authors

In Advances in Neural Information Processing Systems (NeurIPS), 2026

Bib PDF

@inproceedings{piras2026latent,
  title = {Latent-space Attacks for Refusal Evasion in Language Models},
  author = {Piras, Giorgio and Mura, Raffaele and Brau, Fabio and Pintor, Maura and Oneto, Luca and Roli, Fabio and Biggio, Battista},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2026},
  url = {https://openreview.net/forum?id=DWoSECPSPI},
}

AAAI

SOM Directions Are Better than One: Multi-Directional Refusal Suppression in Language Models

Giorgio Piras, Raffaele Mura, Fabio Brau, and 3 more authors

Proceedings of the AAAI Conference on Artificial Intelligence, Mar 2026

DOI Bib PDF Poster

@article{piras2026som,
  title = {SOM Directions Are Better than One: Multi-Directional Refusal Suppression in Language Models},
  volume = {40},
  url = {https://ojs.aaai.org/index.php/AAAI/article/view/40551},
  doi = {10.1609/aaai.v40i39.40551},
  number = {39},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  author = {Piras, Giorgio and Mura, Raffaele and Brau, Fabio and Oneto, Luca and Roli, Fabio and Biggio, Battista},
  year = {2026},
  month = mar,
  pages = {32728-32736},
}

CVPR

1-Lipschitz Layers Compared: Memory Speed and Certifiable Robustness

Bernd Prach, Fabio Brau, Giorgio Buttazzo, and 1 more author

In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun 2024

Bib PDF Code Poster

@inproceedings{prach20241,
  author = {Prach, Bernd and Brau, Fabio and Buttazzo, Giorgio and Lampert, Christoph H.},
  title = {1-Lipschitz Layers Compared: Memory Speed and Certifiable Robustness},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = jun,
  year = {2024},
  pages = {24574-24583},
}

AAAI

Robust-by-Design Classification via Unitary-Gradient Neural Networks

Fabio Brau, Giulio Rossolini, Alessandro Biondi, and 1 more author

Proceedings of the AAAI Conference on Artificial Intelligence, Jun 2023

Bib PDF

@article{brau2023robust,
  title = {Robust-by-Design Classification via Unitary-Gradient Neural Networks},
  volume = {37},
  number = {12},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  author = {Brau, Fabio and Rossolini, Giulio and Biondi, Alessandro and Buttazzo, Giorgio},
  year = {2023},
  month = jun,
  pages = {14729-14737},
}

TPAMI

On the Minimal Adversarial Perturbation for Deep Neural Networks With Provable Estimation Error

Fabio Brau, Giulio Rossolini, Alessandro Biondi, and 1 more author

IEEE Transactions on Pattern Analysis and Machine Intelligence, Jun 2023

Bib PDF

@article{brau2023minimal,
  author = {Brau, Fabio and Rossolini, Giulio and Biondi, Alessandro and Buttazzo, Giorgio},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  title = {On the Minimal Adversarial Perturbation for Deep Neural Networks With Provable Estimation Error},
  year = {2023},
  volume = {45},
  number = {4},
  pages = {5038-5052},
  keywords = {Perturbation methods;Robustness;Estimation;Neural networks;Deep learning;Error analysis;Computational modeling;Adversarial robustness;deep neural networks;trustworthy AI;verification methods},
}