CyLab Seminar: Paulo Esteves-Veríssimo
September 27, 2024
12:00 p.m. ET
Zoom or Newell-Simon Hall 4305
*Please note: this CyLab seminar is open only to partners and Carnegie Mellon University faculty, students, and staff.
Speaker:
Paulo Esteves-Veríssimo
Professor of Computer Science, Director of the Resilient Computing and Cybersecurity Center (RC3)
King Abdullah University of Science and Technology (KAUST)
Talk Title:
Safe and Secure AI/ML-driven Autonomous Vehicles? Not anywhere near yet...
Abstract:
Currently, there is a substantial push towards autonomous vehicles (AV) in the market. However, autonomous vehicles, though using extensive fault-tolerance e.g., in x-by-wire functions, are still not quite safe from an accidental faults’ perspective, leave alone malicious faults.
The growing number of recent AV architectures hinging on the advancements in AI/ML models,
given their functional power, must be put in context with an equally significant number of related
serious or fatal accidents. I advance by presenting important misconceptions and pitfalls
originating from the use of inappropriate AI/ML techniques in the AV area, which may be the
cause of serious accidents.
Furthermore, AV present an even greater threat surface to combined accidental faults and
malicious attacks. These pitfalls have been very slowly recognized by car makers, with
potentially harming results.
However, simultaneously securing unavoidable properties of Safety and Security is indeed a
hard problem. I raise a bit of the curtain on how to break this chicken and egg dilemma---
closing the safety-security gap--- presenting some recent solution avenues based on cyber
resilience, a core subject of my research.
Namely, I will discuss how in-car architectures might evolve to achieve fault and intrusion
resilience of ECUs at system level, as well as reconciling the data-level stochastic nature of
AI/ML paradigms with the determinism of driving control theory at system level, leveraging the
best from both worlds: trustworthiness and intelligence.
Bio:
Paulo Esteves-Veríssimo is a professor of Computer Science at KAUST, founding director of the Resilient Computing and Cybersecurity Center (RC3- https://rc3.kaust.edu.sa/), and research fellow at the University of Luxembourg (LU). He is Fellow of IEEE and of ACM, author of over 200 peer-refereed publications and co-author of 5 books. He is currently interested in resilient computing, in areas like: SDN-based infrastructures; autonomous vehicles; distributed control systems; digital health and genomics; or blockchain and cryptocurrencies.