Individualized Cybersecurity Research Mentoring (iMentor) Workshop 2023

Title: Rethinking Online Trust in the AI Era

Abstract: Large Language Models (LLMs) are powerful tools for generating context-specific high-quality text. However, their potential misuse in fraudulent activities, scams, or for political interference is a concern. Identifying such content is challenging, especially when these models are also being used to enhance legitimate and benign communications.

In this presentation, I will discuss a new cryptographic design that brings “accountability” to direct communication applications, hence building up more trust. First, I will revisit some of the assumptions the solutions from literature made in the context of phishing and spam and how they will change. Then, I will detail the design and discuss the guarantees it can provide. On the way, I will gently touch on how to stitch multiple ideas borrowed from previous research results in a new application context.

Bio: I am a senior researcher at Microsoft Research (MSR) at Redmond. I am trained in applied cryptography. In my 2 years at MSR, I have researched on how to bring more privacy to the end users without harming security of the platforms. Recently, I have started thinking about such privacy-security trade-offs that emerge in new technologies. More specifically, I am interested in preventing disruptions in communication due to Large Language Models.