Deepfake Video & Audio Detection Using Spatial-Temporal CNN-LSTM Networks
Automated forensic detection tool to distinguish synthetic manipulated face videos using ResNet and Bidirectional LSTM.
Problem Statement
The surge in hyper-realistic AI deepfakes poses severe cybersecurity, political, and identity fraud threats, rendering traditional media verification methods obsolete.
Proposed Solution
A dual-stream spatial-temporal deep learning network that extracts per-frame facial artifacts using ResNet-50 while capturing temporal eye-blink and lip-motion inconsistencies using a Bidirectional LSTM.
Key System Features
System Architecture & Pipeline
End-to-end data transformation pipeline from input capture to output visualization.
Project Modules Breakdown
Video Preprocessing & Cropping
Module 1Breaks uploaded video into sequential frames and applies MTCNN to isolate facial bounding boxes.
Spatial Feature Extractor (ResNet-50)
Module 2Computes 2048-dimensional convolutional features representing textural irregularities per frame.
Temporal Sequence Analyzer (Bi-LSTM)
Module 3Processes 30-frame sequence vectors to detect unnatural inter-frame optical flow changes.
Classification & Reporting Dashboard
Module 4Softmax classifier predicting Real vs Fake probability alongside a downloadable PDF audit.
Complete Technology Stack
Viva-Voce Questions Preview
View 50+ Viva Q&A →Sample questions external college examiners ask during project reviews, with model answers:
Package Deliverables
We don't just send files; we explain the code line-by-line so you pass your review with flying colors.
Similar Projects You Might Like
AI-Based Smart Attendance System Using Real-Time Facial Recognition
Automated contactless attendance tracking system with anti-spoofing liveness detection and automated Excel reporting.
Generative AI Medical Document Summarizer & Clinical Diagnostic Assistant (RAG)
Healthcare Retrieval-Augmented Generation (RAG) assistant using LangChain, Vector DB, and LLMs for doctor report synthesis.
Cloud-Native Hospital Management & Tele-Consultation Platform (MERN Stack)
Production-grade multi-tenant healthcare SaaS with doctor scheduling, digital prescriptions, Razorpay, and WebRTC video consultation.