Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020097182 - PRIVACY-PRESERVING VISUAL RECOGNITION VIA ADVERSARIAL LEARNING

Publication Number WO/2020/097182
Publication Date 14.05.2020
International Application No. PCT/US2019/060037
International Filing Date 06.11.2019
IPC
G06F 21/62 2013.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
21Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
60Protecting data
62Protecting access to data via a platform, e.g. using keys or access control rules
G06T 9/00 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
9Image coding
G06N 3/08 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
CPC
G06F 21/6245
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
21Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
60Protecting data
62Protecting access to data via a platform, e.g. using keys or access control rules
6218to a system of files or objects, e.g. local or distributed file system or database
6245Protecting personal data, e.g. for financial or medical purposes
G06N 3/0454
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0454using a combination of multiple neural nets
G06N 3/08
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
Applicants
  • NEC LABORATORIES AMERICA, INC. [US]/[US]
Inventors
  • SOHN, Kihyuk
  • CHANDRAKER, Manmohan
  • TSAI, Yi-Hsuan
Agents
  • BITETTO, James J.
Priority Data
16/674,42505.11.2019US
62/756,76507.11.2018US
62/878,78626.07.2019US
Publication Language English (EN)
Filing Language English (EN)
Designated States
Title
(EN) PRIVACY-PRESERVING VISUAL RECOGNITION VIA ADVERSARIAL LEARNING
(FR) RECONNAISSANCE VISUELLE PRÉSERVANT LA CONFIDENTIALITÉ PAR L'INTERMÉDIAIRE DE L’APPRENTISSAGE CONTRADICTOIRE
Abstract
(EN)
A method for protecting visual private data by preventing data reconstruction from latent representations of deep networks is presented. The method includes obtaining latent features (316) from an input image (312) and learning, via an adversarial reconstruction learning framework (318), privacy -preserving feature representations to maintain utility performance and prevent the data reconstruction by: simulating a black-box model inversion attack by training a decoder (332) to reconstruct the input image from the latent features and training an encoder (314) to maximize a reconstruction error to prevent the decoder from inverting the latent features while minimizing the task loss.
(FR)
L'invention concerne un procédé de protection de données privées visuelles par prévention de la reconstruction de données à partir de représentations latentes de réseaux profonds. Le procédé comprend l'obtention de caractéristiques latentes (316) à partir d'une image d'entrée (312) et l'apprentissage, par l'intermédiaire d'un cadre d'apprentissage de la reconstruction contradictoire (318), des représentations de caractéristiques de préservation de la confidentialité pour maintenir des performances de service public et empêcher la reconstruction de données par : simulation d'une attaque d'inversion de modèle de boîte noire par formation d'un décodeur (332) pour reconstruire l'image d'entrée à partir des caractéristiques latentes et entraîner un codeur (314) pour maximiser une erreur de reconstruction pour empêcher le décodeur d'inverser les caractéristiques latentes tout en réduisant au minimum la perte de tâche.
Also published as
Latest bibliographic data on file with the International Bureau