Keynotes
Trustworthy AI... for Systems Security
Lorenzo Cavallaro, University College London
Abstract:
No day goes by without reading machine learning (ML) success stories across various application areas. Systems security is no exception, where ML's tantalizing performance leave one to wonder whether there are any unsolved problems left. However, machine learning has no real clairvoyant abilities and once the magic wears off, we're left in uncharted territory. Is machine learning truly capable of ensuring systems security? In this talk, we will highlight the importance of reasoning beyond mere performance by examining the consequences of adversarial attacks and distribution shifts in realistic settings. When relevant, we will also delve into behind-the-scenes aspects to encourage reflection on the reproducibility crisis. Our goal is to foster a deeper understanding of machine learning's role in systems security and its potential for future advancements.
Bio: Lorenzo Cavallaro is a Full Professor of Computer Science at University College London (UCL), where he leads the Systems Security Research Lab. He grew up on pizza, spaghetti, and Phrack, and soon developed a passion for underground and academic research. Lorenzo's research vision is to enhance the effectiveness of machine learning for systems security in adversarial settings. He works with his team to investigate the interplay between program analysis abstractions, representations, and ML models, and their crucial role in creating Trustworthy AI for Systems Security. Despite his love for food, Lorenzo finds his Flow in science, music, and family.
No day goes by without reading machine learning (ML) success stories across various application areas. Systems security is no exception, where ML's tantalizing performance leave one to wonder whether there are any unsolved problems left. However, machine learning has no real clairvoyant abilities and once the magic wears off, we're left in uncharted territory. Is machine learning truly capable of ensuring systems security? In this talk, we will highlight the importance of reasoning beyond mere performance by examining the consequences of adversarial attacks and distribution shifts in realistic settings. When relevant, we will also delve into behind-the-scenes aspects to encourage reflection on the reproducibility crisis. Our goal is to foster a deeper understanding of machine learning's role in systems security and its potential for future advancements.
Bio: Lorenzo Cavallaro is a Full Professor of Computer Science at University College London (UCL), where he leads the Systems Security Research Lab. He grew up on pizza, spaghetti, and Phrack, and soon developed a passion for underground and academic research. Lorenzo's research vision is to enhance the effectiveness of machine learning for systems security in adversarial settings. He works with his team to investigate the interplay between program analysis abstractions, representations, and ML models, and their crucial role in creating Trustworthy AI for Systems Security. Despite his love for food, Lorenzo finds his Flow in science, music, and family.
Adversarial Prompting: Return of the Adversarial Example
Eric Wong, University of Pennsylvania
Abstract:
Nearly two decades have passed since early adversarial examples were created to attack spam filtering. Researchers have since questioned their practicality and usefulness as a real-world security threat in deep learning. What role can adversarial machine learning take today? In this talk, we will discuss an emerging threat called adversarial prompting: a real-world attack vector for prompt-based generative AI. Adversarial prompts are currently an active security threat with the potential to circumvent safeguards intended to sanitize generated outputs. This talk will explore threat models and black box adversarial attacks for the prompting setting, demonstrating how it can be used to induce targeted behaviors in generative AI.
Bio: Eric Wong is an Assistant Professor in the Department of Computer and Information Science at the University of Pennsylvania. His work focuses on the foundations of robust systems, building on elements of machine learning and optimization to debug, understand, and develop reliable systems. He is a 2020 Siebel Scholar and received an honorable mention for his thesis on the robustness of deep networks to adversarial examples at Carnegie Mellon University advised by Zico Kolter. Prior to joining UPenn, he was a postdoc at CSAIL MIT advised by Aleksander Madry.
Bio: Eric Wong is an Assistant Professor in the Department of Computer and Information Science at the University of Pennsylvania. His work focuses on the foundations of robust systems, building on elements of machine learning and optimization to debug, understand, and develop reliable systems. He is a 2020 Siebel Scholar and received an honorable mention for his thesis on the robustness of deep networks to adversarial examples at Carnegie Mellon University advised by Zico Kolter. Prior to joining UPenn, he was a postdoc at CSAIL MIT advised by Aleksander Madry.
