Hybrid Learning Techniques for Image Forensics and Privacy Protection in the Face of Deep Fake Threats

Abstract

This study investigates the detection of deepfake images and videos on social media platforms such as Instagram
for forensic analysis using hybrid-learning approaches. It highlights the critical importance of safeguarding privacy and
authenticity in digital media. The background draws attention to the growing threat posed by deepfakes, which pose
significant challenges across multiple domains, such as politics and entertainment. Existing methods often depend on visual
features specific to a dataset and struggle to generalize across different manipulation techniques. Moreover, most
approaches focus exclusively on either temporal or spatial features, which limits their capacity to identify complex anomalies
involving fused facial features like the mouth, nose and eyes. Important solutions to these challenges include Convolutional
Neural Network (CNN), Recurrent Neural Networks (RNN) and hybrid architectures that simultaneously capture spatial and
temporal information in deepfake content, such as Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM),
Gated Recurrent Unit (GRU) and Vision Transformers (ViT). Additionally, this paper introduces a novel combination of artifact
inspection and facial landmark recognition to enhance detection accuracy and employs Gated Recurrent Units (GRUs) and
Vision Transformers (ViT) for data augmentation thereby improving model robustness. The effectiveness of the proposed
approach is validated through experiments demonstrating substantially improved deduction accuracy, with improvement
exceeding 1.5% across multiple datasets. However, several challenges remain, including limited robustness to noise, difficulty
in detecting deepfakes in compressed video formats, and dataset imbalances issues. The proposed enhanced hybrid model
exhibits superior detection performance while maintaining adaptability across multiple datasets. Future research will focus
strengthening model generalization to effectively counter emerging deepfake generation techniques.