Development Of Artificial Intelligence-Based Model for Forensic Analysis of Cross-Platform Deepfakes

Abstract

Synthetic media, a product of advancements in artificial intelligence (AI) and machine learning, represents a transformative innovation that reshapes how content is created, manipulated and consumed. One of the most advanced technology in synthetic media is deepfakes. The misuse of deepfakes poses significant threats to privacy, security and societal trust. Cybercriminals exploit this technology for phishing scams, identity theft and spreading misinformation. This study developed an AI-based hybrid model for forensic analysis of cross platform deep fakes to address these gaps. The study developed a hybrid model that integrates 2D CNN, 1D CNN and RNN framework capable of isolating platform-specific spatial and temporal features for forensic analysis of cross-platform deepfakes. The platforms include social media such as Facebook, Instagram, Youtube, Tiktok and Twiter. The research used Celeb DF dataset for training, validation and testing of the hybrid model. Result from the evaluation metrics showed that the model achieved an accuracy of 99 % on the training data and 93.5 % on the test data. Result equally showed that precision, recall and MCC value were 96.8 %, 90 % and 87.2 % respectively on the test data. The model outperformed single CNN and RNN models and some hybrid models reported. This study demonstrated a reliable hybrid model with high detection accuracy for the forensic analysis of deepfakes. Hence, the model has the potential to address the problem of misinformation caused by synthetic media.