An Intellectual Deep Fake Video Detection using Feature Integration Approach with Adaptive and Hybrid Deep Learning

• R Sathish,
• Daniel Madan Raja S,
• E.Arul,

Vol. 20, No. 1, January 31, 2026
10.3837/tiis.2026.01.015, Download Paper (Free):

• Fake Video Detection; Adaptive Hybrid 1D/3D Multi-dilated Convolution-based Residual Attention Network; Improved Clouded Leopard Optimization; Region of Interest; Viola Jones; Texture Features; Motion Features

Abstract

Currently, the deep fake mechanism has been widely used among individuals with smartphone devices. Firstly, the necessary videos are obtained from the standard resources and forwarded to the pre-processing stage. From the input video frames, the face is detected using Viola Jones and then the Region of Interest (ROI) is extracted for the subsequent process. In addition, the pre-processed video frames are used to extract the motion and texture features. After that, the extracted features are fed into the Adaptive Hybrid (1D/3D) Multi-dilated Convolution-based Residual Attention Network (AHMCRAN) for the recognition of deep fake. Here, the AHMCRAN network parameters are optimally determined by the Improved Clouded Leopard Optimization (ICLO) for increasing the performance of deep fake detection. To verify its effectiveness, the implemented deep fake detection model is evaluated by comparing it to existing algorithms and detection approaches.

Statistics

Cite this article