Forged Real Images Throughout History (FRITH)

📢 Please cite the following publication if you use the FRITH dataset:

Khurshid Asghar, Xianfang Sun, Paul L. Rosin, Mubbashar Saddique, Muhammad Hussain, *Zulfiqar Habib (2019), “Edge-Texture Feature based Image Forgery Detection with Cross Dataset Evaluation”, Machine Vision and Applications, pp: 1243–1262, Vol: 30, Issue: 7-8, Standard: 0932-8092 (Print) 1432-1769 (Online), http://doi.org/10.1007/s00138-019-01048-2.

FRITH contains 255 professionally tampered historic images 🖼️ and 155 authentic (untampered) versions.
It includes various forgery types like:

✂️ Copy–move and splicing
🧪 Fake object insertion
🧩 Object manipulation
🎨 Post-processing (lighting, enhancements)

📥 Download ZIP (17MB): Link 1, Link 2

COMSATS Structured Video Tampering Evaluation Dataset (CSVTED)

📢 Please cite the following publication if you use CSVTED:
N. Akhtar, M. Saddique, P.L. Rosin, X. Sun, M. Hussain, Z. Habib, "A Three-Level Benchmark Dataset for Spatial and Temporal Forensic Analysis of Videos", Machine Vision and Applications, vol. 36, no. 86, pp. 1-22, 2025. DOI: https://doi.org/10.1007/s00138-025-01704-w

🔍 CSVTED: A New Benchmark Dataset for Video Forgery Detection

We are pleased to release CSVTED, a three-level dataset structured by tampering quality and video complexity, designed to support research in video forensic analysis.

📦 Dataset Overview:

1047 videos: 133 original + 914 tampered
Captured using multiple cameras
Lighting conditions: morning, noon, evening, night, fog
Includes various tampering types:

 ▫ Frame duplication
▫ Deletion
▫ Insertion
▫ Copy-move
▫ Splicing
▫ Event-Object-Person (EOP) based tampering

⚙ Realistic Tampering:
We used SSIM and Optical Flow to select optimal positions for duplication, insertion, or deletion. This helps maintain visual consistency and avoids abrupt changes, especially by matching object motion direction.

🎞 Formats: avi, mp4, mov
🎯 Suitable for testing both spatial and temporal forgery detection methods.

📥 Download the dataset (41.9 GB): Link 1, Link 2