FAKE-ID: Project Description

Deepfakes are image and video forgeries produced using artificial intelligence (AI). They are primarily used to spread (political) disinformation, but also to manipulate digital proof of identity. With the expansion of digital identification procedures, incidences of deepfake-based fraud are increasingly targeting private individuals.

The FAKE-ID project is developing legal and ethical standards that can be used to identify and classify manipulated images and video data streams. In doing so, it lays the foundations for the development of a detection program that can in turn use AI to assess the (presumed) authenticity of digital images and videos or their forgery.

This FÖPS sub-project deals with the legal, ethical and societal requirements for the development of AI-based detection programs used by police. Among other topics, this project addresses ethical issues in the use of machine learning necessary for the development of the system; transparency, traceability, and responsibilities of decisions made with the help of “weak AI”; legal and ethical requirements for the usability of detection results in legal proceedings; and protection against exclusively automated decisions, as provided for in Art. 22 of the EU Data Protection Regulation. To this end, FÖPS is working closely with governmental and private information technology institutions to design the AI-based detector from the ground up, per the principles of Privacy by Design and by Default in a privacy-compliant manner. In the spirit of participatory research, FÖPS is also seeking to create an in-depth discussion with non-governmental organizations (NGOs) that are focused on the dangers of AI use in a security context.

Partner:

Bundesdruckerei GmbH (Consortium leadership)

Otto-von-Guericke-Universität Magdeburg

Fraunhofer Heinrich Hertz Institut (HHI)

BioID GmbH (BIOID)

Associated: Zentralstelle für Informationstechnik im Sicherheitsbereich (ZITiS), Landeskriminalämter von Berlin und Sachsen-Anhalt, Deutsche Post DHL Group (DP), Cybersec Verbund Sachsen-Anhalt sowie Fachhochschule für Verwaltung und Dienstleistung (FHVD)

Funded by: Federal Ministry of Education and Research, “Artificial Intelligence in Security Research“
Project duration: 1. May 2021 to 31. October 2024