The Bright and Dark Sides of Computer Vision:
Challenges and Opportunities for Privacy and Security
(CV-COPS 2020)

SEC, Glasgow— Friday August 28, 2020

Important Information on Event

The interactive time slots of the workshop will be 8am to 10am and 8pm to 10pm (UTC+1). Video material is available on the ECCV conference site and will not be played during the interactive slot. The interactive sessions will be dedicated to QA.

While QA will happen during the interactive session, we also provide a mean to gather questions ahead of time via the following shared document:
https://docs.google.com/document/d/1uMM3uyTZeYLws4qc76oJXmwSOItx_oW-_DzqH5w1tzY/edit?usp=sharing
The questions will be read during the interactive session.
We are looking forward to seeing you at the workshop.

1st QA session 8am to 10am UTC+1

Invited talks:

In Search of Lost Performance in Privacy-Preserving Deep Learning, Reza Shokri (National University of Singapore) [Video]
Don't Use Computer Vision For Web Security, Florian Tramer (Stanford University) [Video]

Papers:

Black-Box Face Recovery from Identity Features [Video]
Adversarial Training against Location-Optimized Adversarial Patches [Video]
Face-Image Source Generator Identification
Spatio-Temporal Handwriting Imitation [Video]
Privacy-Aware Face Recognition with Lensless Multi-Pinhole Camera [Video]
Dynamic Backdoor Attacks Against Machine Learning Models
Beyond the Spectrum: Fake Image Detection with Hierarchical Structural Artifacts

2nd QA session 8pm to 10pm UTC+1

Invited talks:

Responsible Computing Research Directions for Computer Vision, Asia J. Biega (Microsoft Research) [Video]

Papers:

Black-Box Face Recovery from Identity Features [Video]
Revisiting the Threat Space for Vision-based Keystroke Inference Attacks [Video]
Body Shape Privacy in Images: Understanding Privacy and Preventing Automatic Shape Extraction
Adversarial Training against Location-Optimized Adversarial Patches [Video]
Frequency-Tuned Universal Adversarial Perturbations [Video]
Beyond the Spectrum: Fake Image Detection with Hierarchical Structural Artifacts

Overview

Computer vision is finally working in the real world, but what are the consequences on privacy and security? For example, recent work shows that vision algorithms can spy on smartphone keypresses from meters away, steal information from inside homes via hacked cameras, exploit social media to de-anonymize blurred faces, and reconstruct images from features like SIFT. Vision could also enhance privacy and security, for example through assistive devices for people with disabilities, phishing detection techniques that incorporate visual features, and image forensic tools. Some technologies present both challenges and opportunities: biometrics techniques could enhance security but may be spoofed, while surveillance systems enhance safety but create potential for abuse.

We need to understand the potential threats and opportunities of vision to avoid creating detrimental societal effects and/or facing public backlash. Following up on last year's very successful workshops at CVPR 2017, CVPR 2018, and CVPR 2019 this workshop will continue to explore the intersection between computer vision and security and privacy to address these issues.

Call for Papers and Extended Abstracts

We welcome original research papers and extended abstracts on topics including, but not limited to:

Computer vision-based security and privacy attacks
Biometric spoofing, defenses and liveness detection
Impact of ubiquitous cameras on society
Captchas and other visual Turing tests for online security
Privacy of visual data
Privacy-preserving visual features and representations
Reversibility of image transformations
Secure/encrypted computer vision and image processing
Wearable camera privacy
Attacks against computer vision systems
Copyright violation detection
Counterfeit and forgery detection
Privacy implications of large-scale visual social media
Other relevant topics

Research papers should contain original, unpublished research, and be up to 14 pages (excluding references). Research papers will be published in the ECCV Workshop Proceedings.

Extended abstracts about preliminary, ongoing or published work should be up to 2 pages (including references). Extended abstracts will be published and archived on this website.

All submissions should be anonymized and will undergo double-blind peer review. Papers and abstracts must be formatted according to the ECCV guidelines and submitted via CMT website. Accepted submissions will be invited for oral or poster presentation at the workshop.

We have two rounds of submission:
1st round:

Submission: May 1st
Notification: May 29th

2nd round:

Submission: July 12th
Notification: July 29th

Invited Speakers

Florian Tramer
Stanford University

Reza Shokri
National University of Singapore

Asia J. Biega
Microsoft Research

Invited talks

Don't Use Computer Vision For Web Security, Florian Tramer (Stanford University) [Video]
In Search of Lost Performance in Privacy-Preserving Deep Learning, Reza Shokri (National University of Singapore) [Video]
Responsible Computing Research Directions for Computer Vision, Asia J. Biega (Microsoft Research) [Video]

Full papers

Black-Box Face Recovery from Identity Features [Video]
Razzhigaev, Anton; Tursynbek, Nurislam; Petiushko, Aleksandr; Kireev, Klim; Kaziakmedov, Edgarl
Revisiting the Threat Space for Vision-based Keystroke Inference Attacks [Video]
Lim, John; Price, True; Monrose, Fabian; Frahm, Jan-Michael
Body Shape Privacy in Images: Understanding Privacy and Preventing Automatic Shape Extraction
Sattar, Hosnieh; Krombholz, Katharina ; Pons-Moll, Gerard; Fritz, Mario
Adversarial Training against Location-Optimized Adversarial Patches [Video]
Rao, Sukrut; Stutz, David; Schiele, Bernt
Frequency-Tuned Universal Adversarial Perturbations [Video]
Deng, Yingpeng; Karam, Lina
Face-Image Source Generator Identification
Salamah, Mohamad; Hel-Or, Hagit
Spatio-Temporal Handwriting Imitation [Video]
Mayr, Martin; Stumpf, Martin; Nikolaou, Anguelos; Seuret, Mathias; Maier, Andreas K; Christlein, Vincent
Privacy-Aware Face Recognition with Lensless Multi-Pinhole Camera [Video]
Ishii, Yasunori; Yamashita, Takayoshi; Sato, Satoshi

Extended papers

Dynamic Backdoor Attacks Against Machine Learning Models
Salem, Ahmed; Wen, Rui; Backes, Michael; Ma, Shiqing; Zhang, Yang
Beyond the Spectrum: Fake Image Detection with Hierarchical Structural Artifacts
He, Yang; Yu, Ning; Keuper, Margret; Fritz, Mario