Capstone
Cybersecurity
Network Intrusion Detection System
Machine-learning anomaly detection over live network traffic
- Python
- Scapy
- XGBoost
- Kafka
- +2
₹5,499₹9,499
4.9/97 reviews
Cybersecurity · Capstone
Deepfake Video Detection
Frame CNN plus a temporal model that flags manipulated faces
5-6 weeks build
Overview
Detects face-swap and reenactment videos by cropping faces from each frame with a CNN feature extractor, then reading the sequence with a temporal model that catches the flicker and blending artefacts a still image hides. It is trained on the public FaceForensics++ dataset, and the report presents ROC and per-manipulation metrics rather than a headline accuracy figure.
What makes it stand up
- 01Face cropping and alignment before any classification
- 02Frame-level CNN features fused by an LSTM over time
- 03Trained and evaluated on public FaceForensics++ data
- 04ROC, AUC and per-manipulation-type breakdown in the report
- 05Web tool that returns a per-clip probability with a heatmap
Module breakdown
- 01
Preprocessing
Detects, crops and aligns faces from sampled frames.
- 02
Frame model
CNN that scores each face crop for manipulation cues.
- 03
Temporal model
Sequence model that reads inconsistency across frames.
- 04
Evaluation
Reports AUC and metrics split by manipulation method.
- 05
Demo app
Uploads a clip and shows a verdict with a saliency heatmap.
After this, you will be able to
- Explain the visual artefacts deepfake detectors rely on
- Defend combining spatial and temporal models over frames alone
- Demo detection on held-out FaceForensics++ clips
Technology stack
- Python
- PyTorch
- OpenCV
- MediaPipe
- Flask
- Level
- Capstone
- Domain
- Cybersecurity
- Build time
- 5-6 weeks
- Difficulty
You receive
- Complete, commented source code
- 60-90 page project report (IEEE format)
- Editable presentation deck
- Architecture, ER and UML diagrams
- Local setup and deployment guide
- Sample dataset or seed data
- Viva question bank with answers
- One 45-minute walkthrough call
- 30 days of doubt-clearing support