MPI-SP Mini Security and Privacy Symposium

 

Abstracts

Don’t shout “Bingo!” Understanding (and Addressing) the Shortcomings of Enterprise Threat Detection Products

We are still awful at preventing data breaches and other cybersecurity incidents. Why are these sophisticated (and costly) commercial threat detection products continuing to fail? In this talk, I'll describe our efforts to better understand, and even address, these failure points. First, I'll provide evidence that the extraordinarily high false alarm rates observed in Endpoint Detection & Response (EDR) products can be eliminated by examining the history of alert-triggering processes. Second, I'll explain how the metrics used to evaluate threat detection products often paint a deeply misleading picture of organizations' security readiness. I will conclude by discussing how our ongoing work seeks to resolve industry shortcomings by providing more principled foundations for threat detection and assessment.

 

Automatically synthesizing exploits for code injection attacks in Node.js packages using types

JavaScript is widely used in applications, via the Node.js runtime. Unfortunately, code-injection vulnerabilities, such as those that allow arbitrary code execution, are frequently found in Node.js packages. Most existing vulnerability-detection tools don’t construct proof-of-concept (PoC) exploits to help developers investigate potential vulnerabilities. In this talk, I will present our work on automatically generating exploits to help developers understand code-injection vulnerabilities. Node.js packages do not come with type specifications, which makes it challenging to generate appropriate inputs to APIs. Our novel exploit synthesis algorithms use output from program analysis to generate inputs of a specific type or with a specific structure and non-trivial interactions with the package API under test. Our evaluations show that we can synthesize PoC for a large fraction of the reported code injection vulnerabilities.

 

On the importance of understanding the user in the age of AI

The increasing reliance on AI brings about new risks. In this talk I'll describe several projects on AI risks that are united by a theme: that understanding the human in the loop -- whether the attacker, the defender, or the potential victim -- can help us better assess and mitigate risks. I'll show how separating attacker goals from strategy can lead to attacks that more effectively fool object recognition (ICML 2022). I'll describe how instantiating abstract attacker goals with more concrete ones can lead to new definitions of risk, and how these, in turn, enable the creation of both stronger attacks and better defenses (NDSS 2024). Finally, I'll discuss some risks of generative AI, namely how prompt suggestions can be used to mislead chatbot users (CHI 2025) and how to improve assessments of LLM answer quality (EMNLP 2025). 

 

Speakers

Limin Jia is a Research Professor of Electrical and Computer Engineering Department at Carnegie Mellon University and a member of CyLab, Carnegie Mellon's computer security and privacy institute. She received her Ph.D. from Princeton in 2008. Her research is in the intersection of programming languages, formal methods, and computer security. She is particularly interested in applying formal methods to analyzing the security guarantees of software systems and to developing mechanisms to make software systems more secure.

Adam Bates is an Associate Professor at the University of Illinois at Urbana-Champaign, where he studies a broad range of topics in computer security. He is best known for his work on data provenance, the practice of examining suspicious activities on computing systems based on their historical context. Fittingly, Adam also appreciates the historical context of computer security research, regularly forcing students in his courses to read James Anderson's 1972 Computer Security Technology and Planning Study… both volumes. Adam is the recipient of two distinguished paper awards (S&P'23, ESORICS'22) and was the runner-up for the ACM SIGSAC Dissertation Award. His research has been recognized and supported by an NSF SaTC FRONTIER, NSF CISE Research Initiation Initiative (CRII), and NSF CAREER Awards, as well as a gift from the VMWare University Research Fund.

Lujo Bauer is a Professor of Electrical and Computer Engineering, and of Computer Science, at Carnegie Mellon University. He is also a member of CyLab, Carnegie Mellon's computer security and privacy institute. He received his B.S. in Computer Science from Yale University in 1997 and his Ph.D., also in Computer Science, from Princeton University in 2003. Lujo served as the program (co-)chair for the flagship computer security conferences of the IEEE (S&P 2015), the Internet Society (NDSS 2014), and USENIX (Security 2025). Lujo's research examines many aspects of computer security and privacy, and balances attention to the human users of systems with attention to software and algorithms. His current research topics include: studying the risks raised by our increasing reliance on AI; using AI to make us more secure, including by finding software bugs faster and by detecting attacks; and developing tools to make smart homes safer for all users.