Using the CNN Architecture Based on the EfficientNetB4 Model to Efficiently Detect Deepfake Images

Arman Zhalgasbayev, Tolegen Aiteni, Nursultan Khaimuldin

2024

Closed access Metadata (CSL-JSON)
DOI 10.1109/ieeeconf61558.2024.10585385
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

This study evaluates the effectiveness of using the EfficientNetB4 model architecture for deepfake image detection and identifies the main challenges in developing an accurate model for deepfake image detection. Deepfake image detection is a complex task that requires lots of data, advanced models and new approaches for handling features in face images which shows that image is real or fake. This research is based on the Deepfake Detection Challenge (DFDC), which took place in 2019–2020, where world experts showed new approaches to solving this problem. During the research and development of the model, the best practices and experience of experts were used to identify the deepfake images.