ICCV 2023 Workshop on Adversarial RObustness in the Real World (AROW)

Welcome to the 4th Workshop on Adversarial Robustness In the Real World (AROW)

ICCV 2023, Paris

Recent deep-learning-based methods achieve great performance on various vision applications. However, insufficient robustness on adversarial cases limits real-world applications of deep-learning- based methods. AROW workshop aims to explore adversarial examples, as well as, evaluate and improve the adversarial robustness of computer vision systems.

News & Updates


Computer vision systems achieve advanced performance, however, research in adversarial machine learning shows concurrent vision systems are less robust in comparison to human vision systems. Perturbation-based adversarial examples achieve a significant impact on vision systems accuracy but remain a gap toward real-world scenarios. While recent works demonstrate deep-learning-based methods are also vulnerable to those more real-world adversarial examples, e.g., partial occlusions, atmospheric changes, and style changes. Such vulnerability potentially limits the real-world usage of computer vision systems. Discovering, evaluating, and defending those real-world adversarial examples helps to understand and improve the robustness of computer vision systems, which can help the deployment of computer vision systems in safety-critical applications. This workshop aims to bring together researchers from various fields, including adversarial machine learning, robust vision, and explainable AI, to discuss recent research and future directions for adversarial robustness and explainability, with a particular focus on real-world scenarios.

The topics involved in the workshop include but are not limited to:


Liwei Wang

Peking University

Yinpeng Dong

Tsinghua University

Yixuan Li

University of Wisconsin-Madison

Florian Tramèr

ETH Zürich

Matthias Hein

University of Tübingen

Yisen Wang

Peking University

Schedule (Oct. 2)

The timezone is the local timezone of the ICCV23 conference.

Session Time Speaker Content
Morning 9:00 - 9:10 N/A Welcome and Introduction
9:10 - 9:40 Mikel Rodriguez Invited Talk #1: Some lessons learned from observing real-world attacks against vision systems
9:40 - 10:10 Sharon Yixuan Li Invited Talk #2: TBD
10:10 - 10:40 Matthias Hein Invited Talk #3: TBD
10:40 - 11:00 N/A Coffe Break
11:00 - 11:30 Yinpeng Dong Invited Talk #4: On Robustness of Vision Models in the 3D World
Poster Session 11:30 - 12:30 N/A Poster Session
12:30 - 13:30 N/A Lunch Break
Afternoon 13:30 - 14:00 Liwei Wang Invited Talk #5: Robust Generalization Requires Exponentially Large Neural Networks
14:00 - 14:30 Yisen Wang Invited Talk #6: Rethinking Adversarial Examples (tentative)
14:30 - 15:00 Florian Tramer Invited Talk #7: Where are the real-world attacks?
15:00 - 15:30 Caner Hazirbas Invited Talk #8: Unveiling Model Vulnerabilities with the Recent Robustness Datasets
15:30 - 16:00 N/A Coffe Break
Best Paper Session 16:00 - 17:00 N/A Best Paper Session
17:00 - 17:40 N/A Closing Remarks

Organizing Committee

Yutong Bai

Johns Hopkins University

Mo Zhou

Johns Hopkins University

Angtian Wang

Johns Hopkins University

Vikash Sehwag

Princeton University

Adam Kortylewski

Johns Hopkins University

Cihang Xie

University of California, Santa Cruz

Yisen Wang

Peking University

Xinyun Chen

University of California, Berkeley

Judy Hoffman

Georgia Tech

Dawn Song

University of California, Berkeley

Jun Zhu

Tsinghua University

Rama Chellapa

Johns Hopkins University

Gang Hua

Wormpex AI Research

Vishal M. Patel

Johns Hopkins University

Yinpeng Dong

Tsinghua University

Yingwei Li

Johns Hopkins University

Alan L. Yuille

Johns Hopkins University

Siyue Wang

Microsoft Security Research

Program Committee

Publication Process and Author Guide

Please follow the ICCV2023 author policies, as well as the ICCV2023 submission guidelines. It is also suggested to check the ICCV2023 Author FAQs. We accept two types of submissions, i.e., long papers and extended abstracts.

Long papers in the AROW workshop will be published by IEEE/CVF. The following guidelines must be respected:

We also accept extended abstracts. Note, different from the long papers:

Information for Presenters

  1. Invited talk

    Each workshop room will be equipped with A/V systems for presentation, including a HDMI cable to connect one machine (speaker's PC or Mac), 2 microphones on the podium, 2 wireless microphones, and video projectors. The aspect ratio of the screen will be 16:9. Please encourage presenters to attend the workshop in person and communicate with participants at the venue. If the presenters are not able to attend in person, remote presentation is also allowed. In this case, no registration is needed for invited speakers. Note, however, that remote presenters with IEEE/CVF workshop papers must have (at least one-day) workshops/tutorials pass registration. Please check out the authors note on the registration page: https://iccv2023.thecvf.com/registration-81.php

  2. Poster presentation (95.4 cm x 138.8 cm, portrait format)

    Workshop poster sessions will be held in each workshop room, and poster panels will be provided and arranged inside the room. Note that the poster size for workshops is different from that for the main conference. The workshop poster panel size will be 95.4 cm x 138.8 cm (WxH, aspect ratio 0.69:1, portrait format). A0 paper in portrait would well fit the panel by some margin. Please make sure to notify this information to the workshop poster presenters. There will be an on-site printing service from which you can collect your printed poster. You will receive more information about the on-site printing service in a separate email later.

  3. Workshop streaming

    If you plan to live-stream your workshop, a Wi-Fi network enough for streaming will be provided for each workshop room. However, we do not provide technical support for workshop streaming, and thus streaming of workshops should be managed by each organizer.

Accepted Long Papers

In total 12 long papers are accepted. They are still going through the publication process with IEEE.

Accepted Extended Abstracts

In total 6 extended abstracts are accepted. PDFs are hosted on this website.

Please contact us (see bottom of this page) if you are interested in sponsoring this workshop!

Here are some reference links:

Social Network Updates

📢 [Deadline Extension] Good news! We have extended the submission deadline for the 4th Workshop on Adversarial Robustness In the Real World, ICCV2023!

📅 New DDL: July 20, 2023, 23:59 PT

📷 Workshop Website: https://t.co/r9tu5UvuG8#AROW #ICCV2023 #AdversarialRobustness

— M. Zhou (@MZhou73277685) July 18, 2023

📢 [Call For Papers] We invite participants to submit their work to the 4th Workshop on Adversarial Robustness In the Real World, ICCV 2023, France!

📷 Workshop Website: https://t.co/r9tu5UvuG8#AROW #ICCV2023 #AdversarialRobustness #DeepLearning #ComputerVision #Paris

— M. Zhou (@MZhou73277685) July 2, 2023

📢 Exciting news! Join us at the 4th Workshop on Adversarial Robustness In the Real World, happening at ICCV 2023 in Paris, France. 🌍🤖

🔗 Workshop Website: https://t.co/ELnVUJyg3H#AROW #ICCV2023 #AdversarialRobustness #DeepLearning #ComputerVision #Paris

— M. Zhou (@MZhou73277685) July 2, 2023