I am a Postdoctoral Researcher at the sAIfer lab, in the Department of Electrical and Electronic Engineering at the University of Cagliari (UNICA), Italy. My research focuses on Adversarial Machine Learning and Large Language Models security.
I got my BSc in Electrical, Electronic, and Computer Engineering in July 2019 with mark 110/110, and my MSc with honors in Computer Engineering, Artificial Intelligence and Cybersecurity in July 2021 from the University of Cagliari. In November 2021 I started my PhD at the Sapienza University of Rome in the context of the National PhD-AI program. I obtained my PhD with honors in January 2025 discussing the thesis titled Adversarial Pruning: Improving Evaluations and Methods, supervised by Prof. Battista Biggio.
During my PhD, I have been a visiting student at the Karlsruhe Institute for Technology, Germany, supervised by Prof. Christian Wressnegger. I serve as a reviewer for conferences including NeurIPS, ICLR, AAAI, USENIX, and ACM CCS, and journals including Pattern Recognition, Neurocomputing, Machine Learning, and IEEE TIFS.
This paper uses Self-Organizing Maps to extract multiple refusal directions in language models, showing that multi-directional ablation can suppress refusal more effectively than a single-direction baseline.
This paper introduces LatentBreak, a white-box jailbreak attack that creates natural, low-perplexity adversarial prompts through latent-space feedback.
For a full list of my publications, checkout my Google Scholar profile!