development of a metamodel for detecting ‘deepfakes’

On the occasion of the ‘European Cyber ​​Week’, which takes place from November 19 to 21 in Rennes, whose theme is that of artificial intelligence, the teams of Thales participated in the AID (Defense Innovation Agency) Challenge. As part of this challenge relating to the detection of images generated by modern AI platforms, the cortAIx teams, the Thales AI accelerator, have developed a metamodel for detecting images generated by AI (‘deepfakes ‘).

This tool aims to fight against image manipulationfor different use cases, such as the fight against identity fraud.

This Thales metamodel is built on an aggregation of models, each assigning an authenticity score to each image. It draws on ‘machine learning’ techniques, decision trees, and evaluation of the strengths and weaknesses of each model in order to analyze the authenticity of an image.

The ‘Friendly Hackers’ team, at the origin of this invention, is part of cortAIx, Thales’ AI accelerator, which has more than 600 AI researchers and engineers, including 150 based on the Saclay plateau. The Group’s ‘Friendly Hackers’ have developed a toolbox, the ‘BattleBox’, whose objective is to facilitate the assessment of the robustness of systems integrating AI against attacks aimed at exploiting the vulnerabilities of different models AI (including Large Language Models).