Research Topics | Publications | Thesis Topics | Misc

🧠 Research Topics

My research focuses on improving and understanding the trustworthiness of AI architectures. I am particularly interested in the interplay between robustness, security, and efficiency, exploring how we can design AI systems that are both powerful and reliable under real-world constraints. Some of the central questions driving my work include:

  • 🛡️ safety and security AI threats and countermeasures, with a focus on adversarial robustness and out of distribution samples in computer vision and natural language processing.
  • 🔍 Explainability and uncertainty estimation as tools to strengthen trust in AI decisions and to better interpret feature patterns in large and complex vision and NLP models.
  • 🚦 Risk-aware metrics to evaluate failures of AI in safety-critical domains such as autonomous driving or healthcare.

“Any technological advance can be dangerous. Fire was dangerous from the start, and so (even more so) was speech — and both are still dangerous to this day — but human beings would not be human without them.”

— Isaac Asimov, The Caves of Steel

📄 Publications

For a complete and updated list of my publications, please visit: Google Scholar · Semantic Scholar

  • 2025

    1. Exploiting Edge Features for Transferable Adversarial Attacks in Distributed Machine Learning
      G. Rossolini, F.Brau, A. Biondi, B.Biggio, G. Buttazzo
      Internet of Things · Link
    2. Video Deblurring by Sharpness Prior Detection and Edge Information
      Y, Tian, F. Brau, G. Rossolini, G. Buttazzo, H. Meng
      arXiv · Link
    3. Leveraging Knowledge Graphs and LLMs for Structured Generation of Misinformation
      S. Nayab, M. Simoni, G. Rossolini
      20th International Conference on Availability, Reliability and Security (ARES 2025) · Link
    4. SynDRA: Synthetic Dataset for Railway Applications
      G. D'Amico, F. Nesti, G. Rossolini, M. Marinoni, S. Sabina, G. Buttazzo
      Proceedings of the Winter Conference on Applications of Computer Vision (WACV) · Link
    5. Improving LLM Reasoning for Vulnerability Detection via Group Relative Policy Optimization
      M Simoni, A. Fontana, G. Rossolini, A. Saracino
      (preprint) arXiv · Link
    6. GTPO: Trajectory-Based Policy Optimization in Large Language Models
      M Simoni, A. Fontana, G. Rossolini, A. Saracino
      (preprint) arXiv · Link
    7. Benchmarking the Spatial Robustness of DNNs via Natural and Adversarial Localized Corruptions
      G. Marchiori Pietrosanti, G. Rossolini, A. Biondi, G.Buttazzo
      Pattern Recognition 172:112412 · Link

  • 2024

    1. Attention-Based Real-Time Defenses for Physical Adversarial Attacks in Vision Applications
      G. Rossolini, A. Biondi, G. Buttazzo
      2024 ACM/IEEE 15th International Conference on Cyber-Physical Systems (ICCPS) · Link
    2. Towards Trustworthy AI: Understanding the Impact of AI Threats and Countermeasures
      G. Rossolini,
      Società Italiana di Intelligence (SOCINT) - to appear · Link
    3. CARLA-GeAR: A Dataset Generator for a Systematic Evaluation of Adversarial Robustness of Deep Learning Vision Models
      F. Nesti, G. Rossolini, G. D’Amico, A. Biondi, G. Buttazzo
      IEEE Transactions on Intelligent Transportation Systems · Link
    4. Concise Thoughts: Impact of Output Length on LLM Reasoning and Cost
      S. Nayab, G. Rossolini, M. Simoni, A. Saracino, G. Buttazzo, N. Manes, F. Giacomelli
      arXiv · Link
    5. Edge-Only Universal Adversarial Attacks in Distributed Learning
      G. Rossolini, T. Baldi, A. Biondi, G. Buttazzo
      arXiv · Link

  • 2023

    1. On the Minimal Adversarial Perturbation for Deep Neural Networks with Provable Estimation Error
      F. Brau, G. Rossolini, A. Biondi, G. Buttazzo
      IEEE Transactions on Pattern Analysis and Machine Intelligence · Link
    2. Defending From Physically-Realizable Adversarial Attacks Through Internal Over-Activation Analysis
      G. Rossolini, F. Nesti, F. Brau, A. Biondi, G. Buttazzo
      Proceedings of the AAAI Conference on Artificial Intelligence (AAAI2023) · Link
    3. Robust-by-Design Classification via Unitary-Gradient Neural Networks
      F. Brau, G. Rossolini, A. Biondi, G. Buttazzo
      Proceedings of the AAAI Conference on Artificial Intelligence (AAAI2023) · Link
    4. TrainSim: A railway simulation framework for LiDAR and camera dataset generation
      G. D’Amico, M. Marinoni, F. Nesti, G. Rossolini, G. Buttazzo, S. Sabina, G. Lauro
      IEEE Transactions on Intelligent Transportation Systems · Link
    5. On the real-world adversarial robustness of real-time semantic segmentation models for autonomous driving
      G. Rossolini, F. Nesti, G. D’amico, S. Nair, A. Biondi, G. Buttazzo
      IEEE Transactions on Neural Networks and Learning Systems · Link

  • 2022

    1. Increasing the confidence of deep neural networks by coverage analysis
      G. Rossolini, A. Biondi, G. Buttazzo
      IEEE Transactions on Software Engineering · Link
    2. in-Car Entertainment via Group-wise Temporary Mobile Social Networking.
      M. Cimino, A. Di Tecco, P. Foglia, R. Giannessi, J. Malvatani, CA. Prete, G. Rossolini
      VEHITS 2022 · Link
    3. Evaluating the robustness of semantic segmentation for autonomous driving against real-world adversarial patch attacks
      G. Rossolini, F. Nesti, S. Nair, A. Biondi, G. Buttazzo"
      Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision · Link

🎓 Thesis Topics

For available master’s and doctoral thesis projects, please contact me.
Projects span a range of research areas, including (but not limited to) the following:

  • Adversarial Robustness for Vision
    Localized attacks/defenses, universal perturbations, transferability (CNNs vs ViTs).
  • Trustworthy AI for Safety-Critical Systems
    Risk-aware metrics, robustness–accuracy trade-offs, runtime monitoring.
  • Distributed & Split Inference (Edge–Cloud)
    Offloading reliability, early exits, edge-only attacks/defenses, feature-space security.
  • LLM Reliability & Efficiency
    Reasoning under constraints, robust prompting, and LLM alignment.
  • Out-of-Distribution & Uncertainty
    OOD detection & calibration under distribution shift.
  • Dataset Tools & Benchmarks
    Spatial robustness benchmarking (natural + adversarial localized corruptions).
  • Explainability for Robust Systems
    Internal activation analysis, attribution-guided defenses, failure mode interpretation.

📌 Misc (Awards, Career & Services)

  • PhD Dissertation Prize — Carlo Mosca Award, Società Italiana di Intelligence (link)
  • Project Coordinator — “On the Safety and Security of Distributed AI-based Autonomous Multi-Agent Systems,” financed by the Department of Excellence in Robotics & AI, Scuola Superiore Sant’Anna, Pisa.
  • Associate Editor for the The Visual Computer journal (since 2024).
  • Consulting Associate Editor for the IEEE Transactions on Information Forensics and Security journal (since 2024).
  • Reviewer / Program Committee Member for several AI conferences and journals — IEEE T-PAMI, IEEE T-IFS, IEEE T-ITS, IEEE T-NNLS, ICCV 2023, ECCV 2024, CVPR 2025, NeurIPS 2025, AAAI 2023–2026, and others.
  • Session Chair — DSD-HSTIEC 2024 and 2025.

📬 Contact

giulio.rossolini@santannapisa.it

Last update

July, 2025