SCP Security Seminar

We have two talks from students at University of Illinois Urbana-Champaign and University of Chicago! 

Talk #1

Speaker: Jaron Mink

Title: DeepPhish: Understanding User Trust Towards Artificially Generated Profiles in Online Social Networks

Abstract: Machine learning models are becoming increasingly powerful and are actively deployed in security- and privacy-critical systems; however, this shift has inevitably led to abusive machine learning applications which are lucrative for adversaries and harmful to users. Notably, deepfake-generated content has been increasingly used in social profiles to construct artificial personas which serve disinformation or perform social engineering attacks on other users in online social networks. Many of these victims are attempting to understand and navigate these security and privacy threats for the first time, requiring adjustments in their behavior and responsibilities. In this talk, I will discuss how end-users perceive the novel threat of detecting deepfake social profiles from genuine, human-crafted ones. Through this work, we will discuss what implications exist for content moderators, social media platforms, and future defenses.

Biography: Jaron Mink is a PhD Candidate of the Computer Science Department at University of Illinois at Urbana-Champaign. He received his Bachelor’s of Science (Magna Cum Laude) in the field of Computer Science at University of California, Los Angeles in 2019. Jaron investigates computer security and privacy threats and focuses on users’ perception and mitigation of emerging concerns. His work has appeared in venues such as CHI, USENIX Security, IEEE S&P, WWW, and has been reported on by the Scientific American and the 21st Show. He is a recipient of the NSF Graduate Research Fellowship (GRFP). Jaron also serves as a consultant to Partnership on AI, investigating ways to better anticipate AI risks. He has spent two summers working with faculty at the Max Planck Institute for Security and Privacy (2021) and the Max Planck Institute for Software Systems (2022).

Talk #2

Speaker: Shinan Liu

Title: LEAF: Navigating Concept Drift in Cellular Networks

Abstract: Operational networks commonly rely on machine learning models for many tasks, including detecting anomalies, inferring application performance, and forecasting demand. Yet, model accuracy can degrade due to concept drift, whereby the relationship between the features and the target prediction changes. Mitigating concept drift is thus an essential part of operationalizing machine learning models in the context of networking---or regression models in general. Unfortunately, as we show, concept drift cannot be sufficiently mitigated by frequently retraining models using newly available data, and doing so can even degrade model accuracy further. In this paper, we characterize concept drift in a large cellular network for a major metropolitan area in the United States. We find that concept drift occurs across many important key performance indicators (KPIs), independently of the model, training set size, and time interval---thus necessitating practical approaches to detect, explain, and mitigate it. To do so, we develop Local Error Approximation of Features (LEAF). We introduce LEAF and demonstrate its effectiveness on a variety of KPIs and models. LEAF detects drift; explains features and time intervals that most contribute to drift, and mitigates drift using forgetting and over-sampling. We evaluate LEAF against industry-standard mitigation approaches (notably, periodic retraining) with more than four years of cellular KPI data. Our initial tests with a major cellular provider in the US show that LEAF is effective on complex, real-world data. LEAF consistently outperforms periodic and triggered retraining while reducing costly retraining operations.

Biography: Shinan Liu is a 4th year Ph.D. student in the Computer Science Department at the University of Chicago advised by Prof. Nick Feamster. He is interested in networked systems, security, explainable AI, and measurement. Typical scenarios he has explored include Cellular Networks, Internet of Things, and Cyber Physical Systems. Shinan publishes papers in USENIX Security and his past work was reported by Forbes, The Wall Street Journal, ACM TechNews, and more. Shinan is also the recipient of the Daniels Fellowship.