Umberto Salviati

Umberto Salviati

phd student Robust AI, Univesity of Padua

Umberto

Umberto Salviati

Ph.D. Student in Robust AI & Machine Learning Security
University of Padua and FBK

Short Bio

Ph.D. candidate in Computer Science at the University of Padua, enrolled in the BMCS (Brain, Mind and Computer Science) Ph.D. program within the Department of Mathematics. My research specializes in Robust AI and Machine Learning Security in adversarial settings. I focus on enhancing the reliability and trustworthiness of AI systems deployed in sensitive and high-risk environments, with particular emphasis on emerging vulnerabilities in NLP pipelines and LLM-based architectures.

Current Position

  • Ph.D. Student — BMCS Ph.D. Course, Department of Mathematics, University of Padua
    Research Area: Robust AI and Machine Learning Security — FBK (Fondazione Bruno Kessler), Augmented Intelligence Center

Education

  • Ph.D. in Computer Science (Ongoing) — BMCS Ph.D. Program, Department of Mathematics, University of Padua
    Focus: Robust AI and Machine Learning Security
  • M.Sc. in Computer Science — University of Padua
    Thesis: * Thesis: *Autonomous Driving on Mars: From Dataset to Models - A Deep Learning Application on Martian Imagery**

Publications

  • Matteo Gioele Collu, Umberto Salviati, Roberto Confalonieri, Mauro Conti, Giovanni Apruzzese — Publish to Perish: Prompt Injection Attacks on LLM-Assisted Peer Review, arXiv:2508.20863 (2025).
    Study on prompt injection vulnerabilities in LLM-driven peer review systems.
  • Alberto Castagnaro, Umberto Salviati, Mauro Conti, Luca Pajola, Simeone Pizzi — The Hidden Threat in Plain Text: Attacking RAG Data Loaders, arXiv:2507.05093 (2025).
    Analysis of security weaknesses in document ingestion pipelines of Retrieval-Augmented Generation systems.

Full list available on Google Scholar.

Teaching Experience

2025–2026

  • Cybersecurity and Cryptography: Principles and Practices – Course page
  • Computer Architecture (Architettura degli Elaboratori) – Course page

2024–2025

  • Machine Learning for Finance – Course page
  • Cybersecurity and Cryptography: Principles and Practices (SCQ0089579) – Course page

Research Interests

  • Robust Machine Learning
  • ML Security & Adversarial Attacks
  • Retrieval-Augmented Generation Security
  • Natural Language Processing & Deep Learning

Collaboration & Contact

I am open to collaborations in the areas of ML security, robust AI, NLP, and adversarial machine learning.
If interested in joint research or academic exchange, feel free to get in touch:

📧 Email: umberto.salviati@phd.unipd.it
🏫 University page
🏢 FBK research Center 🧠
🔗 Google Scholar

Thank you for your interest in my academic profile.