Detect Watermark Tampering
Analyze if an image has been tampered with or if the watermark has been removed.
Upload Image to Analyze
Drop an image for forensic analysis
or click to browse
Full Analysis: Combines forensic methods with ML detection (200-500ms).
Fast Forensics: Rule-based detection only, instant results (100-300ms).
Tampering Analysis Results
Hybrid Forensic Analysis System
Image tampering leaves behind invisible mathematical scars. Our system combines traditional signal processing with modern machine learning to expose these anomalies.
Signal Processing Layer
- Laplacian Edge Detection: When an object is pasted from another image, the edges often have mismatched sharpness. We analyze the variance of the Laplacian text to spot these inconsistencies.
- Error Level Analysis (ELA): Different parts of an image compressed at different times show different error rates. ELA highlights "newer" pixels (edits) in bright colors against the background.
- Frequency Analysis (FFT): Copy-move forgery often creates periodic patterns in the frequency domain. We use Fast Fourier Transform to spot these unnatural repetitions.
AI Classification Layer
We pass the results of the signal processing layer into a lightweight Convolutional Neural Network (CNN). The AI looks for high-level semantic inconsistencies that math alone might miss.
- Authentic (Green): Signal is consistent. No obvious edits.
- Suspicious (Red/Yellow): Conflicting noise patterns or double-compression artifacts detected. Manual review advised.