DETECTION ALGORITHMS AND MODELS FOR SYNTHETICALLY GENERATED VISUAL MEDIA

The rapid development of artificial intelligence technologies has unprecedentedly expanded the capabilities for the automatic generation and manipulation of visual content. This, in turn, has led to a proliferation of deepfakes and other artificially generated objects that pose a significant threat to the reliability of digital information. This paper analyzes improved algorithms and models that enable the high-accuracy detection of artificially generated visual objects. Within the framework of this study, in contrast to traditional approaches that rely solely on spatial features, a hybrid architecture is proposed that evaluates temporal discontinuities and logical inconsistencies between consecutive video frames. The developed model was tested on prominent open-source datasets, and the results confirmed its superior effectiveness in detecting artificial manipulations compared to baseline methods. The proposed algorithmic solution can be practically implemented to enhance digital security and verify information authenticity across various information systems.

Hali tarjima qilinmagan