LLM Security Day

About this event

This event is organized as an on-site event only.

LLM Security Day

1. Talk: "Trustworthy AI: How to Prevent Your Machine Learning Models From Becoming a Security Liability"
By: Tanja Šarčević and Anastasia Pustozerova
2. Talk: "Using LLMs for Offensive Security"
By: Andreas Happe
Short Discussion

Talk language: English
On-site event only!: Floragasse 7, 1040 Vienna (5. floor)
Further information will be announced soon!

Agenda
*********
17:55: Gathering
18:05: Welcome
18:10: Talk: "Trustworthy AI: How to Prevent Your Machine Learning Models From Becoming a Security Liability"
18:40: Talk:"Using LLMs for Offensive Security"
19:10 Discussion
20:00: Food, Drinks & Get together

Looking forward to seeing you there!

Talks & Speaker Details:

"Using LLMs for Offensive Security"

Since 2023, researchers have used LLMs to drive offensive security, e.g., pen-testing. This talk will highlight academic research on LLM-driven pen-testing between 2023--2025, show (the speaker's) current prototypes/research, and detail current engineering/research questions.

Speaker Andreas Happe:
Andreas is a developer gone pen-tester gone phd student that currently focuses on the the intersection between LLMs and Penetration-Testing. Otherwise, you can spot him failing at local bouldering gymns.

Trustworthy AI: How to Prevent Your Machine Learning Models From Becoming a Security Liability

As AI systems power more products, decisions, and customer experiences, one uncomfortable truth is emerging: machine learning introduces entirely new privacy and security risks that most organizations are not prepared for. Models can leak sensitive data, be reverse-engineered, manipulated, or compromised. This talk demystifies what can go wrong when deploying AI in real-world environments and shows how to build trustworthy, resilient, and regulation-ready ML systems.

Speaker: Anastasia Pustozerova
Anastasia received her bachelor’s degree in Applied Mathematics & Physics from St. Petersburg University. She has a joint master’s degree from a European Master’s Program in Computational Logic (TU Wien, TU Dresden, Free University of Bozen-Bolzano). Anastasia is currently working towards a PhD. She is involved in an European research project FeatureCloud as an expert in federated machine learning, where she is working on cyber risk assessment and mitigation.

Speaker: Tanja Šarčević
Tanja received a bachelor´s degree in Computer Science from the University of Zagreb and a master´s degree in Logic and Computation at the TU Wien. She is currently working towards her PhD degree with the focus on ownership protection of data and machine learning models.