| Vision | We want to gain a deeper understanding of membership inference attacks (MIAs) that are prevalent in LLMs. This will allow us to demonstrate the privacy risks and how they can leak private data. |
| Mission | Our team will solve this problem by designing, implementing, and documenting a functional proof-of-concept Membership Inference Attack. We will reproduce existing MIA code to test on a specified open-source LLM, and document the process and results. |
| Team Members | Owen Walpole, Andrew Stilgenbauer, and Sean Phelps |
| Client | Dr. Honglu Jiang |
| Site | N/A |