Deep Natural Anonymization

Certification: 03/2022

Important: The certificate with the „Cert. No.EP-P-8NSBJG according to version v201701 has expired on 31.03.2024“. Brighter AI is currently in the certification process according to Article 42 GDPR for „processors“, the certification process should be completed by Q3 2024.

DNAT (Deep Natural Anonymization) is a redaction feature for images and videos. It automatically detects personal identifiers (faces and license plates) and generates synthetic replacements reflecting the original attributes. Therefore, DNAT facilitates protecting identities while keeping necessary information for analytics or machine learning.

Deep Natural Anonymization Technology (DNAT)
(a feature of the software brighter Redact Enterprise v3)

Qualification: IT product

View the DNAT certificate

Cert. No.

EP-P-8NSBJG

Version of Certification Criteria

01/2017

Validity

22/03/2022 – 31/03/2024

Public Report

Certification 2022: DNAT Short Public Report [PDF] 

Manufacturer/Provider

brighter AI Technologies GmbH

Litfaß-Platz 2
10178 Berlin
Germany

BEST

DNAT is solely used for detecting faces and license plates in images and videos as well as generating synthetic replacements so that industry-leading recognition algorithms are not able to identify the original face or license plate anymore. In cases where realistic human face attributes and license plate layouts are necessary for analytical or software development purposes and machine learning, the quality of the images remains, but the original faces and license plates are not part of the redacted footage anymore. Hereby, DNAT clearly facilitates the anonymization of images and videos.

For the redaction of license plates DNAT provides the option for the customer to use individual and customized output layer. This feature ensures that a synthetic license plate will not exist in reality by accident.

ATTENTION

Depending on the redacted images and videos, the use of DNAT may not always result in a complete anonymization of the respective footage. E.g., if a person wears a name tag, it may be possible to identify them even if their image is detected and redacted by the use of DNAT.

Customers should make use of the feature ensuring that a synthetic license plate will not exist in reality by accident (cf. above at BEST).

SUMMARY

Brighter Redact Enterprise is a software for the redaction of images and videos, which is to be hosted on customer’s premise. The software provides for several redaction features, including precision blur, selective redaction and deep natural anonymization (DNAT). Only the latter feature is covered by the current certification. DNAT automatically detects faces and license plates in footage and generates synthetic replacements reflecting the original attributes. Therefore, it facilitates protecting identities while keeping necessary information for analytics or machine learning.

DETAILS

Initial Cert 202203

Target of Evaluation (ToE) is DNAT, a feature of the software brighter Redact Enterprise v3. The ToE includes the following components:

  • Microservices
    • Face / license plate detection (deep learning model)
    • Face / license plate tracking
    • Face attributes (deep learning model)
    • Face position map (deep learning model)
    • License plate landmarks (deep learning model)
    • Face / license plate filter
    • Face occlusion (deep learning model)
    • Face / license plate landmarks smoothing
    • Face / license plate generator
    • License plate optical reconstruction
    • Image composer
    • Video aggregator
    • Data deletion (garbage collector)
  • Interfaces
    • Mounted drive to docker container (API – input file)
    • Internal between the DNAT microservices (Kafka)
    • Docker container to mounted drive (API – output file, analytics file)

The ToE does not include the following:

  • Hard- and software components
    • Video splitter in brighter Redact Enterprise
    • NVIDIA Toolkit (processing data from a docker container with GPU)
    • Numpy (library to process data with GPU)
    • Host machine, Ubuntu operating system, mounted drive and GPU
    • Data storage outside brighter Redact Enterprise
    • Job related meta data
    • Original video system (original video / image file production)
    • Data storage outside the host machine
    • Analytics software
    • Network and transport components outside brighter Redact Enterprise
    • Other functionalities / features than DNAT forming part of brighter Redact Enterprise
    • License Management
  • Services
    • DNAT provided as SaaS
    • Other services than DNAT
  • Interfaces
    • Docker to external elastic search server (https – frame-based billing)
    • GUI (BAI, Flassger)

Legal Evaluator

Prof. Dr. Ralf B. Abel
Oktaviostr. 129
22043 Hamburg
Germany

Technical Evaluator

Marc Neumann
IBS data protection services and consulting GmbH
Zirkusweg 1

Deep Natural Anonymization

Certification: 03/2022

DNAT (Deep Natural Anonymization) is a redaction feature for images and videos. It automatically detects personal identifiers (faces and license plates) and generates synthetic replacements reflecting the original attributes. Therefore, DNAT facilitates protecting identities while keeping necessary information for analytics or machine learning.

Deep Natural Anonymization Technology (DNAT)
(a feature of the software brighter Redact Enterprise v3)

Qualification: IT product

View the DNAT certificate

Cert. No.

EP-P-8NSBJG

Version of Certification Criteria

01/2017

Validity

22/03/2022 – 31/03/2024

Public Report

Certification 2022: DNAT Short Public Report [PDF] 

Manufacturer/Provider

brighter AI Technologies GmbH

Litfaß-Platz 2
10178 Berlin
Germany

BEST

DNAT is solely used for detecting faces and license plates in images and videos as well as generating synthetic replacements so that industry-leading recognition algorithms are not able to identify the original face or license plate anymore. In cases where realistic human face attributes and license plate layouts are necessary for analytical or software development purposes and machine learning, the quality of the images remains, but the original faces and license plates are not part of the redacted footage anymore. Hereby, DNAT clearly facilitates the anonymization of images and videos.

For the redaction of license plates DNAT provides the option for the customer to use individual and customized output layer. This feature ensures that a synthetic license plate will not exist in reality by accident.

ATTENTION

Depending on the redacted images and videos, the use of DNAT may not always result in a complete anonymization of the respective footage. E.g., if a person wears a name tag, it may be possible to identify them even if their image is detected and redacted by the use of DNAT.

Customers should make use of the feature ensuring that a synthetic license plate will not exist in reality by accident (cf. above at BEST).

SUMMARY

Brighter Redact Enterprise is a software for the redaction of images and videos, which is to be hosted on customer’s premise. The software provides for several redaction features, including precision blur, selective redaction and deep natural anonymization (DNAT). Only the latter feature is covered by the current certification. DNAT automatically detects faces and license plates in footage and generates synthetic replacements reflecting the original attributes. Therefore, it facilitates protecting identities while keeping necessary information for analytics or machine learning.

DETAILS

Initial Cert 202203

Target of Evaluation (ToE) is DNAT, a feature of the software brighter Redact Enterprise v3. The ToE includes the following components:

  • Microservices
    • Face / license plate detection (deep learning model)
    • Face / license plate tracking
    • Face attributes (deep learning model)
    • Face position map (deep learning model)
    • License plate landmarks (deep learning model)
    • Face / license plate filter
    • Face occlusion (deep learning model)
    • Face / license plate landmarks smoothing
    • Face / license plate generator
    • License plate optical reconstruction
    • Image composer
    • Video aggregator
    • Data deletion (garbage collector)
  • Interfaces
    • Mounted drive to docker container (API – input file)
    • Internal between the DNAT microservices (Kafka)
    • Docker container to mounted drive (API – output file, analytics file)

The ToE does not include the following:

  • Hard- and software components
    • Video splitter in brighter Redact Enterprise
    • NVIDIA Toolkit (processing data from a docker container with GPU)
    • Numpy (library to process data with GPU)
    • Host machine, Ubuntu operating system, mounted drive and GPU
    • Data storage outside brighter Redact Enterprise
    • Job related meta data
    • Original video system (original video / image file production)
    • Data storage outside the host machine
    • Analytics software
    • Network and transport components outside brighter Redact Enterprise
    • Other functionalities / features than DNAT forming part of brighter Redact Enterprise
    • License Management
  • Services
    • DNAT provided as SaaS
    • Other services than DNAT
  • Interfaces
    • Docker to external elastic search server (https – frame-based billing)
    • GUI (BAI, Flassger)

Legal Evaluator

Prof. Dr. Ralf B. Abel
Oktaviostr. 129
22043 Hamburg
Germany

Technical Evaluator

Marc Neumann
IBS data protection services and consulting GmbH
Zirkusweg 1

EuroPriSe

Die deutschlandweit erste Zertifizierung für Auftragsverarbeiter nach Artikel 42 DSGVO.

Über EuroPriSe
Kontakt

Joseph-Schumpeter-Allee 25
53227 Bonn

EuroPriSe

Die deutschlandweit erste Zertifizierung für Auftragsverarbeiter nach Artikel 42 DSGVO.

Kontakt

Joseph-Schumpeter-Allee 25
53227 Bonn

© All Rights Reserved.

© All Rights Reserved.