Status:

• Researcher at RSAC--"Semi-honest-model-unbeliever" seeking security and privacy measures that transcend trust, ideal assumptions, and heuristic security definitions (but, in reality, I spend most of my time either breaking ML models or building ML models to break stuff).
• Working on:
  • Security and Privacy in Machine Learning:
  • Collaborative Learning
  • LLMs (yes, I know..)
  • Password Security
  • Security Cryptosystems (when someone asks me to)

Preprints:

• Dario Pasquini, Evgenios M. Kornaropoulos, Giuseppe Ateniese.
  Hacking Back the AI-Hacker: Prompt Injection as a Defense Against LLM-driven Cyberattacks
  Media coverage: Dark Reading, Schneier, The Stack
• Dario Pasquini, Evgenios M. Kornaropoulos, Giuseppe Ateniese.
  LLMmap: Fingerprinting For Large Language Models

Works I like to talk about:

• [ACM AISec 24] Dario Pasquini, Martin Strohmeier, Carmela Troncoso.
  Neural Exec: Learning (and Learning from) Execution Triggers for Prompt Injection Attacks
• [IEEE S&P'24] -- Dario Pasquini, Danilo Francati, Giuseppe Ateniese, Evgenios M. Kornaropoulos.
  Breach Extraction Attacks: Exposing and Addressing the Leakage in Second Generation Compromised Credential Checking Services
  
• [IEEE S&P'24] -- Dario Pasquini, Giuseppe Ateniese, Carmela Troncoso.
  Universal Neural-Cracking-Machines: Self-Configurable Password Models from Auxiliary Data
• [IEEE S&P'23] -- Dario Pasquini, Mathilde Raynal, Carmela Troncoso.
  On the (In)security of Peer-to-Peer Decentralized Machine Learning
• [ACM CCS'22] -- Dario Pasquini, Danilo Francati, Giuseppe Ateniese.
  Eluding Secure Aggregation in Federated Learning via Model Inconsistency
• [ACM CCS'21] -- Dario Pasquini, Giuseppe Ateniese, Massimo Bernaschi.
  Unleashing the Tiger: Inference Attacks on Split Learning
• [USENIX Sec'21] -- Dario Pasquini, Marco Cianfriglia, Giuseppe Ateniese, Massimo Bernaschi.
  Reducing Bias in Modeling Real-world Password Strength via Deep Learning and Dynamic Dictionaries
• [IEEE S&P'21] -- Dario Pasquini, Ankit Gangwal, Giuseppe Ateniese, Massimo Bernaschi, Mauro Conti.
  Improving Password Guessing via Representation Learning

Talks (the ones I remembered to put here):

• "Neural Exec: Learning Execution Triggers for Prompt Injection Attacks" [Swiss AI Sec. Network, Sapienza University]
• "Privacy Preserving Collaborative Machine Learning?" [CNR Italy, Inria Rennes, Monash University, Roma3 University, JPMorgan AlgoCRYPT Center of Excellence]
• "Password Alchemy with Universal Neural-Cracking-Machines: Transmuting email addresses into Password Models" [PasswordsCon'23, Roma3 University]

Program Committees:

• ACM CCS 2023, USENIX Sec 2023/2025, IEEE SaTML 2024/2025, PETS 2025.

Things I don't work on, but I'd love to 🔮:

• Hardware sec, Cryptanalysis.