ATVA 2021 Workshop on Security and Reliability of Machine Learning (SRML)
Date: Oct 18th, 2021
Location: The workshop will be held virtually
Zoom link: https://cmu.zoom.us/j/97234065013?pwd=VExidmdpaHJWODFLVEFjVGZEeUlXQT09
Registration link: https://formal-analysis.com/atva/2021/workshop.html
As machine learning models have increasingly been deployed in real-world applications, concerns are raised about the dangers of manipulation and misuse of these models, especially in security-critical domains. To analyze the reliability of the machine learning systems under malicious settings, traditional empirical attacks can easily provide an empirical upper bound of robustness performance. On the other hand, many recent certification techniques have been proposed and achieved great success in providing formal guarantees of the reliability of the machine learning systems, gradually approaching the certified lower bound of the robustness performance towards the empirical upper bounds.
This workshop will focus on recent research and future directions about the security and reliability of machine learning systems, from both empirical and certifiable analysis perspectives. We aim to bring together experts from machine learning, security, and formal verification communities highlighting recent work. Hopefully, we can chart out important and promising future directions towards secure and reliable machine learning systems and encourage cross-community collaborations.
Recorded Workshop Video: https://www.youtube.com/watch?v=XIRQ9WXT0y0
Schedule - Time Zone: Pacific Time (PT)
You can use this tool to convert time to your time zone.
8:45 am -9:00 am PT: Opening remarks
Session 1:
9:00 am - 9:40 am PT: Keynote talk from David Wagner (UC Berkeley)
Title: Defense against adversarial examples
10:00 am -10:20 am PT: Contributed talk 1 from Patrick Henriksen (Imperial College London)
Title: VeriNet: A Symbolic Interval Propagation-Based Approach for Formal Verification of Neural Networks
10:20 am -10:40 am PT: Contributed talk 2 from Maksym Andriushchenko (École Polytechnique Fédérale de Lausanne)
Title: RobustBench: a Standardized Adversarial Robustness Benchmark
10:40 am -11:00 am PT: Contributed talk 3 from Alessandro De Palma (University of Oxford)
Title: Improving Branch and Bound for Neural Network Verification
11:00 am -11:20 am PT: Contributed talk 4 from Krishnamurthy (Dj) Dvijotham (DeepMind)
Title: Verification beyond Lp norm robustness: Probabilistic Specifications and Global Robustness
11:20 am -11:40 am PT: Contributed talk 5 from Aditi Raghunathan (Stanford University)
Title: The Many Facets of Robust Machine Learning: From Mathematical Guarantees to Real-world Shifts
11:40 am -12:30 pm PT: Panel session 1:
Matthias Hein (University of Tübingen), Cho-Jui Hsieh (UCLA), Taylor T. Johnson (Vanderbilt University), Wan-Yi Lin (Bosch Center for AI), Aditi Raghunathan (Stanford University), Florian Tramèr (Stanford University)
Session 2:
1:00 pm -1:40 pm PT: Keynote talk from Zico Kolter (Carnegie Mellon University)
Title: Robustness Beyond the Worst (or the Average) Case
2:00 pm -2:20 pm PT: Contributed talk 6 from Greg Yang (Microsoft)
Title: Randomized Smoothing of All Shapes and Sizes
2:20 pm -2:40 pm PT: Contributed talk 7 from Yizheng Chen (Columbia University)
Title: Learning Security Classifiers with Verified Global Robustness Properties
2:40 pm -3:00 pm PT: Contributed talk 8 from Hoang-Dung Tran (University of Nebraska, Lincoln)
Title: Robustness Verification for Semantic Segmentation Networks using Relaxed Reachability
3:00 pm -3:20 pm PT: Contributed talk 9 from Hadi Salman (Massachusetts Institute of Technology)
Title: Certified Patch Robustness via Smoothed Vision Transformers
3:20 pm -3:40 pm PT: Contributed talk 10 from Wenbo Guo (Pennsylvania State University)
Title: Explaining and Remediating Deep Reinforcement Learning Policies
3:40 pm -4:30 pm PT: Panel session 2:
Stanley Bak (Stony Brook University), Yizheng Chen (Columbia University), Nicholas Carlini (Google Brain), Changliu Liu (Carnegie Mellon University), David Wagner (UC Berkeley), Eric Wong (MIT)
Invited Speakers
Keynote
Panel
Stony Brook University
Bosch Center for AI
University of California, Los Angeles
Vanderbilt University
University of California, Berkeley
Carnegie Mellon University
Stanford University
MIT
Columbia University
Talk
Imperial College London
University of Oxford
DeepMind
Stanford University
Microsoft
Columbia University
University of Nebraska, Lincoln
MIT
Pennsylvania State University
Organizers
Columbia University
Carnegie Mellon University
Drexel University
Columbia University