Abstract :

Today, Deepfake Media Is Able To Create Faces And Movements With Increasing Realism, Which Poses A Threat To Digital Trust, Identity Security, Journalism And Cyber Forensics. In This Research, A Detection Method Based On Spatial And Temporal Analysis For Authenticating Pictures And Videos Is Presented. The Work Process Is Divided Into Media Upload, Frame Extraction, Face Localisation, Resizing And Normalisation, Deep Feature Extraction And Binary Real/fake Classification. The Materials Of The Work Describe CNN, Vision Transformer And LSTM Based Analysis. The Benchmark Study Which Accompanies It Demonstrates The Combination Of Enhanced EfficientNet-B0 And Temporal Convolutional Network Incorporating Multi-scale Feature Fusion. The Reported Baseline Achieves 91.5% Training Accuracy And 92.45% Testing Accuracy On The Balanced FFIW-10K Dataset After 40 Epochs With The Computation Cost Of 0.45 GFLOPs. Realistic Media Upload And Confidence, Frame By Frame Analysis With A Flask-based Interface. Based On The Findings From Multiple Sources, It Can Be Concluded That Multi-feature Spatial-temporal Analysis Can Be A Good Approach That Is Robust For Deepfake Detection And Deployable On Consumer-grade Hardware.

Published:

01-8-2026

Issue:

Vol. 26 No. 8 (2026)

Page Nos:

1139-1145

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite