CONCEPTUAL APPROACH TO DETECTING DEEPFAKE MODIFICATIONS OF BIOMETRIC IMAGES USING NEURAL NETWORKS

Abstract

The National Cybersecurity Cluster of Ukraine is functionally oriented towards building systems to protect various platforms of information infrastructure including the creation of secure technologies for detecting deepfake modifications of biometric images based on neural networks in cyberspace. This space proposes a conceptual approach to detecting deepfake modifications which is deployed based on the functioning of a convolutional neural network and the classifier algorithm for biometric images structured as 'sensitivity-Yuden index-optimal threshold-specificity'. An analytical security structure for neural network information technologies is presented based on a multi-level model of 'resources-systems-processes-networks-management' according to the concept of 'object-threat-defense'. The core of the IT security structure is the integrity of the neural network system for detecting deepfake modifications of biometric face images as well as data analysis systems implementing the information process of 'video file segmentation into frames-feature detection processing - classifier image accuracy assessment'. A constructive algorithm for detecting deepfake modifications of biometric images has been developed: splitting the video file of biometric images into frames - recognition by the detector - reproduction of normalized facial images - processing by neural network tools - feature matrix computation - image classifier construction. Keywords: biometric image deepfake modifications neural network technology convolutional neural network classification decision support system conceptual approach analytical security structure.