Fooling Computer Vision Classifiers with Adversarial Examples

Award Information
Agency: Department of Defense
Branch: Navy
Contract: N68335-18-C-0804
Agency Tracking Number: N182-127-0173
Amount: $124,999.00
Phase: Phase I
Program: SBIR
Solicitation Topic Code: N182-127
Solicitation Number: 2018.2
Timeline
Solicitation Year: 2018
Award Year: 2019
Award Start Date (Proposal Award Date): 2018-10-15
Award End Date (Contract End Date): 2019-04-18
Small Business Information
2501 Earl Rudder Freeway South, College Station, TX, 77845
DUNS: 184758308
HUBZone Owned: N
Woman Owned: N
Socially and Economically Disadvantaged: N
Principal Investigator
 Jason Hill
 (979) 764-2200
 jason.hill@lynntech.com
Business Contact
 Taylor Blincoe
Phone: (979) 764-2200
Email: contract@lynntech.com
Research Institution
N/A
Abstract
The Lynntech team proposes to develop a Computer Vision FoolKit system that integrates cutting-edge approaches to systematically evaluate physically realizable adversarial attacks against several leading computer vision classifiers. It has been noted that most deep neural networks are demonstrably vulnerable to adversarial examples, even in the form of small-magnitude changes in intensities of the input images. Adversarial attacks result in an object being mislabeled from the ground truth class or unlabeled (invisible). Mislabeling attacks can target a particular object class label, or be untargeted resulting in a random class assignment. Physical realizations of such attacks on computer vision systems have been a recent active research topic with much debate about their effectiveness and robustness. Our Computer Vision FoolKit will utilize a general gray-box attack algorithm to take into account numerous physical conditions of the images in order to construct robust adversarial modifications to a targets appearance effective against multiple state-of-the-art classifiers. The FoolKit must incorporate a variety of environmental conditions, including characteristics of images captured from moving platforms, and ultimately extend capabilities across the electromagnetic spectrum. Our FoolKits physical attacks will be straight-forward to construct and deploy and effectively fool real-world vehicle and aircraft classifiers in an operational environment.

* Information listed above is at the time of submission. *

Agency Micro-sites

SBA logo
Department of Agriculture logo
Department of Commerce logo
Department of Defense logo
Department of Education logo
Department of Energy logo
Department of Health and Human Services logo
Department of Homeland Security logo
Department of Transportation logo
Environmental Protection Agency logo
National Aeronautics and Space Administration logo
National Science Foundation logo
US Flag An Official Website of the United States Government