The Secure and Private AI (SPY) Lab conducts research on the security, privacy and trustworthiness of machine learning systems. We often approach these problems from an adversarial perspective, by designing attacks that probe the worst-case performance of a system to ultimately understand and improve its safety.
💡 Learn more about our work and read our publications on our website.
🖥️ Check the code for our projects in this repository.
Finding trojans in aligned LLMs. Official repository for the competition hosted at SaTML 2024.
Code for paper "Universal Jailbreak Backdoors from Poisoned Human Feedback"
Certified robustness "for free" using off-the-shelf diffusion models and classifiers
Repository for the VMI Summer Retreat Workshop on Hacking AI Agents
Official code for "Measuring Non-Adversarial Reproduction of Training Data in Large Language Models" (https://arxiv.org/abs/2411.10242)